Skip to content

Instantly share code, notes, and snippets.

@JackyP

JackyP/Tabular_Neural_Networks_for_Individual_Claims.ipynb Secret

Created March 14, 2022 14:09
Show Gist options
  • Save JackyP/ae6a9fefdb3f0a1aeaf2fb6a1a5ea824 to your computer and use it in GitHub Desktop.
Save JackyP/ae6a9fefdb3f0a1aeaf2fb6a1a5ea824 to your computer and use it in GitHub Desktop.
Download ZIP
Tabular Nets for Individual Claims
Display the source blob
Display the rendered blob
Raw
Tabular_Neural_Networks_for_Individual_Claims.ipynb
{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"id": "Em7RlIuuzk8Q"
},
"source": [
"# Tabular Neural Networks for Individual Claims"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "gzKPDBVF72a8"
},
"source": [
"## Introduction\n",
"\n",
"This is an introductory to intermediate level article on deep learning for claims reserving. The article can be read as a standalone article to learn about pytorch and deep learning, but it follows earlier articles on machine learning for reserving more broadly published on the [Machine Learning for Reserving Working Party Blog](https://institute-and-faculty-of-actuaries.github.io/mlr-blog/).\n",
"\n",
"This article explores deep learning concepts further, introducing several network architectures including Residual Networks, with a focus on \"tabular\" (data table based) architectures. We will see how they perform against the \"LASSO with ramps\" model on our simple synthetic dataset.\n",
"\n",
"The code in this article is fairly generic, and can be adapted easily to other non-reserving regression problems as well.\n",
"\n",
"## Background reading\n",
"Previously, [Self-assembling claim reserving models using the LASSO](https://institute-and-faculty-of-actuaries.github.io/mlr-blog/post/f-lasso/) by Greg Taylor and Grainne McGuire describes a LASSO model with \"ramp\" data transformations used as the benchmark in this article in more detail. \n",
"\n",
"Following that, Grainne McGuire and the author introduced how Python and ``scikit-learn`` can be used [to build machine learning models for actuarial claims reserving using techniques such as decision trees, random forest, gradient boosting, LASSO, and simple neural networks.](https://institute-and-faculty-of-actuaries.github.io/mlr-blog/post/f-scikitexample/). "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "2vfLuFEjAjgC"
},
"source": [
"## Components of a Neural Network"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "8T_Ls-Fz-tDT"
},
"source": [
"A modern deep learning model has a number of key customisable components:\n",
" * **Model architecture**: This is the formula structure, using differentiable functions, that connects the outputs to the inputs and the model weights.\n",
" * **Loss function**: The objective function we want to optimise with respect to model outputs.\n",
" * **Optimizer**: This is how the weights are optimised to reduce the loss.\n",
" * **Initialization strategy**: This is how the weights are initially randomly set, which can affect whether the model gets stuck in local optima, or encounter other problems.\n",
" \n",
"Papers have been written on all of the above to find the \"best\" models from the perspective of accuracy, speed of convergence, memory efficiency and other metrics.\n",
"\n",
"Deep learning model architecture can be quite flexible. For example, deep learning can also produce sequence based models akin to ARIMA where it takes a sequence $y_{n-1}, y_{n-2}...$ to predicting the next element $y_{n}$, such as the [DeepTriangle](https://arxiv.org/abs/1804.09253) model by Kevin Kuo. However, this article will focus on tabular format models, which are similar to decision trees, LASSO, and gradient boosting which take tabular inputs X, and predict a single input y. \n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "iPFFNgfb_9Ek"
},
"source": [
"## Setup"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "UsgmS7oe-vfu"
},
"source": [
"We will set up the deep learning models in [Pytorch](https://pytorch.org/). Pytorch is one of the most popular Python packages for deep learning alongside [Keras](https://keras.io) and [JAX](https://github.com/google/jax), and has been popular particularly for research applications where flexibility and customisability is important. That flexibility is quite helpful for what we want to do in reserving.\n",
"\n",
">*I've been using PyTorch a few months now and I've never felt better. I have more energy. My skin is clearer. My eye sight has improved.* - Andrej Karpathy (2017)\n"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "oDs0lpeU8Zni"
},
"source": [
"The first step is to load all the libraries required for this work (and install them first if needed).\n",
"\n",
"This session uses ``pandas``, ``numpy``, ``scikit-learn``, ``matplotlib``, ``duckdb`` and pytorch (``torch``). The versions are shown below. If you are having problems running our code, check that your packages are the same as ours and ``pip install -U`` upgrade / downgrade them if needed.\n",
"\n",
"This notebook can be run locally or on [Google Colab](https://colab.research.google.com). If you are running this on a local machine and have a GPU you would like to utilise, make sure you [install the right version of pytorch for best performance](https://pytorch.org/get-started/locally/)."
]
},
{
"cell_type": "code",
"execution_count": 249,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "N1avYHra4_NZ",
"outputId": "03b98f82-cc90-4224-df5b-749dad1ca03f"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Name: pandas\n",
"Version: 1.3.5\n",
"Summary: Powerful data structures for data analysis, time series, and statistics\n",
"Home-page: https://pandas.pydata.org\n",
"Author: The Pandas Development Team\n",
"Author-email: pandas-dev@python.org\n",
"License: BSD-3-Clause\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: numpy, python-dateutil, pytz\n",
"Required-by: fastai, seaborn\n",
"---\n",
"Name: numpy\n",
"Version: 1.21.4\n",
"Summary: NumPy is the fundamental package for array computing with Python.\n",
"Home-page: https://www.numpy.org\n",
"Author: Travis E. Oliphant et al.\n",
"Author-email: \n",
"License: BSD\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: \n",
"Required-by: blis, duckdb, fastprogress, h5py, Keras-Preprocessing, lightgbm, matplotlib, opt-einsum, pandas, polars, scikit-learn, scikit-optimize, scipy, seaborn, skorch, spacy, tensorboard, tensorflow-macos, thinc, torchvision, wordcloud\n",
"---\n",
"Name: scikit-learn\n",
"Version: 1.0.2\n",
"Summary: A set of python modules for machine learning and data mining\n",
"Home-page: http://scikit-learn.org\n",
"Author: \n",
"Author-email: \n",
"License: new BSD\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: joblib, numpy, scipy, threadpoolctl\n",
"Required-by: fastai, lightgbm, scikit-optimize, skorch\n",
"---\n",
"Name: matplotlib\n",
"Version: 3.5.1\n",
"Summary: Python plotting package\n",
"Home-page: https://matplotlib.org\n",
"Author: John D. Hunter, Michael Droettboom\n",
"Author-email: matplotlib-users@python.org\n",
"License: PSF\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: cycler, fonttools, kiwisolver, numpy, packaging, pillow, pyparsing, python-dateutil\n",
"Required-by: fastai, seaborn, wordcloud\n",
"---\n",
"Name: torch\n",
"Version: 1.10.1\n",
"Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration\n",
"Home-page: https://pytorch.org/\n",
"Author: PyTorch Team\n",
"Author-email: packages@pytorch.org\n",
"License: BSD-3\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: typing-extensions\n",
"Required-by: fastai, torchaudio, torchvision\n",
"---\n",
"Name: duckdb\n",
"Version: 0.3.2\n",
"Summary: DuckDB embedded database\n",
"Home-page: https://www.duckdb.org\n",
"Author: \n",
"Author-email: \n",
"License: MIT\n",
"Location: /Users/jacky/miniforge3/lib/python3.9/site-packages\n",
"Requires: numpy\n",
"Required-by: \n"
]
}
],
"source": [
"!pip show pandas numpy scikit-learn matplotlib torch duckdb"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "7OtHEPCn-AXr"
},
"source": [
"Import the libraries:"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {
"id": "lLHnqIAg8HZl"
},
"outputs": [
{
"data": {
"text/plain": [
"<duckdb.DuckDBPyConnection at 0x1678e4970>"
]
},
"execution_count": 76,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"\n",
"import duckdb\n",
"\n",
"from sklearn.compose import ColumnTransformer, TransformedTargetRegressor\n",
"from sklearn.preprocessing import OneHotEncoder, StandardScaler, MinMaxScaler\n",
"from sklearn.pipeline import Pipeline\n",
"\n",
"from sklearn.model_selection import GridSearchCV, RandomizedSearchCV\n",
"from sklearn.metrics import mean_squared_error\n",
"\n",
"from sklearn.utils import all_estimators\n",
"\n",
"import matplotlib.pyplot as plt\n",
"from matplotlib.colors import LogNorm\n",
"\n",
"import time\n",
"\n",
"import itertools\n",
"\n",
"from sklearn.base import BaseEstimator, RegressorMixin, TransformerMixin\n",
"from sklearn.utils.validation import check_X_y, check_array, check_is_fitted\n",
"\n",
"import torch\n",
"import torch.nn as nn\n",
"from torch.distributions import Normal, OneHotCategorical\n",
"import torch.nn.functional as F\n",
"\n",
"import math\n",
"\n",
"import random\n",
"random.seed(0) # Set the seed for the random number generation \n",
"np.random.seed(0)\n",
"torch.manual_seed(0)\n",
"\n",
"# start an in-memory database\n",
"con = duckdb.connect(database=':memory:')\n",
"\n",
"# define a fill null function\n",
"con.execute(\"\"\"\n",
" CREATE MACRO FILL_NULL(a, b) AS CASE WHEN a IS NULL THEN b ELSE a END;\n",
"\"\"\")"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "V-GLbRryOrXg"
},
"source": [
"We will also use various helper functions from last time. "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "NyR--K6WBKWc"
},
"source": [
"## Data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Importing Data"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Z0_2mQZF8QTZ"
},
"source": [
"For the reserving data, we use a simulated dataset from the SynthETIC R package. "
]
},
{
"cell_type": "code",
"execution_count": 34,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "Yeb7wG-A8LbE",
"outputId": "db2a3690-e7c7-4ee4-8ac0-e5186af71e99"
},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>pmt_no</th>\n",
" <th>occurrence_period</th>\n",
" <th>occurrence_time</th>\n",
" <th>claim_size</th>\n",
" <th>notidel</th>\n",
" <th>setldel</th>\n",
" <th>payment_time</th>\n",
" <th>payment_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>payment_delay</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>0.623835</td>\n",
" <td>785870.789628</td>\n",
" <td>0.065163</td>\n",
" <td>18.228022</td>\n",
" <td>4.197594</td>\n",
" <td>5</td>\n",
" <td>25104.778182</td>\n",
" <td>25631.935128</td>\n",
" <td>3.508595</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>1</td>\n",
" <td>2</td>\n",
" <td>1</td>\n",
" <td>0.623835</td>\n",
" <td>785870.789628</td>\n",
" <td>0.065163</td>\n",
" <td>18.228022</td>\n",
" <td>7.096012</td>\n",
" <td>8</td>\n",
" <td>26176.620067</td>\n",
" <td>27112.545886</td>\n",
" <td>2.898418</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>1</td>\n",
" <td>3</td>\n",
" <td>1</td>\n",
" <td>0.623835</td>\n",
" <td>785870.789628</td>\n",
" <td>0.065163</td>\n",
" <td>18.228022</td>\n",
" <td>11.157697</td>\n",
" <td>12</td>\n",
" <td>26333.186750</td>\n",
" <td>27828.701791</td>\n",
" <td>4.061685</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>1</td>\n",
" <td>4</td>\n",
" <td>1</td>\n",
" <td>0.623835</td>\n",
" <td>785870.789628</td>\n",
" <td>0.065163</td>\n",
" <td>18.228022</td>\n",
" <td>14.445762</td>\n",
" <td>15</td>\n",
" <td>26341.097381</td>\n",
" <td>28293.903794</td>\n",
" <td>3.288065</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>1</td>\n",
" <td>5</td>\n",
" <td>1</td>\n",
" <td>0.623835</td>\n",
" <td>785870.789628</td>\n",
" <td>0.065163</td>\n",
" <td>18.228022</td>\n",
" <td>18.452453</td>\n",
" <td>19</td>\n",
" <td>592456.913866</td>\n",
" <td>649127.994604</td>\n",
" <td>4.006691</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18978</th>\n",
" <td>3624</td>\n",
" <td>2</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>41.622132</td>\n",
" <td>42</td>\n",
" <td>6586.081338</td>\n",
" <td>8093.128975</td>\n",
" <td>0.670541</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18979</th>\n",
" <td>3624</td>\n",
" <td>3</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>42.081820</td>\n",
" <td>43</td>\n",
" <td>9716.975065</td>\n",
" <td>11967.648057</td>\n",
" <td>0.459688</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18980</th>\n",
" <td>3624</td>\n",
" <td>4</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>42.407479</td>\n",
" <td>43</td>\n",
" <td>7770.338755</td>\n",
" <td>9585.568042</td>\n",
" <td>0.325659</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18981</th>\n",
" <td>3624</td>\n",
" <td>5</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>43.066655</td>\n",
" <td>44</td>\n",
" <td>203618.760893</td>\n",
" <td>252007.199917</td>\n",
" <td>0.659176</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18982</th>\n",
" <td>3624</td>\n",
" <td>6</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>43.354731</td>\n",
" <td>44</td>\n",
" <td>34908.748394</td>\n",
" <td>43266.205665</td>\n",
" <td>0.288075</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>18983 rows × 12 columns</p>\n",
"</div>"
],
"text/plain": [
" claim_no pmt_no occurrence_period occurrence_time claim_size \\\n",
"0 1 1 1 0.623835 785870.789628 \n",
"1 1 2 1 0.623835 785870.789628 \n",
"2 1 3 1 0.623835 785870.789628 \n",
"3 1 4 1 0.623835 785870.789628 \n",
"4 1 5 1 0.623835 785870.789628 \n",
"... ... ... ... ... ... \n",
"18978 3624 2 40 39.767468 270737.291484 \n",
"18979 3624 3 40 39.767468 270737.291484 \n",
"18980 3624 4 40 39.767468 270737.291484 \n",
"18981 3624 5 40 39.767468 270737.291484 \n",
"18982 3624 6 40 39.767468 270737.291484 \n",
"\n",
" notidel setldel payment_time payment_period payment_size \\\n",
"0 0.065163 18.228022 4.197594 5 25104.778182 \n",
"1 0.065163 18.228022 7.096012 8 26176.620067 \n",
"2 0.065163 18.228022 11.157697 12 26333.186750 \n",
"3 0.065163 18.228022 14.445762 15 26341.097381 \n",
"4 0.065163 18.228022 18.452453 19 592456.913866 \n",
"... ... ... ... ... ... \n",
"18978 0.666458 2.920804 41.622132 42 6586.081338 \n",
"18979 0.666458 2.920804 42.081820 43 9716.975065 \n",
"18980 0.666458 2.920804 42.407479 43 7770.338755 \n",
"18981 0.666458 2.920804 43.066655 44 203618.760893 \n",
"18982 0.666458 2.920804 43.354731 44 34908.748394 \n",
"\n",
" payment_inflated payment_delay \n",
"0 25631.935128 3.508595 \n",
"1 27112.545886 2.898418 \n",
"2 27828.701791 4.061685 \n",
"3 28293.903794 3.288065 \n",
"4 649127.994604 4.006691 \n",
"... ... ... \n",
"18978 8093.128975 0.670541 \n",
"18979 11967.648057 0.459688 \n",
"18980 9585.568042 0.325659 \n",
"18981 252007.199917 0.659176 \n",
"18982 43266.205665 0.288075 \n",
"\n",
"[18983 rows x 12 columns]"
]
},
"execution_count": 34,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"transactions = pd.read_csv(\n",
" \"https://raw.githubusercontent.com/JackyP/SyntheticExports/main/synthetic_test_transaction_dataset.csv\"\n",
")\n",
"transactions"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The dataset to use initially is the transaction dataset. It contains paid transactions only.\n",
"\n",
"Id columns:\n",
" * ``claim_no``: Claim no\n",
" * ``pmt_no``: payment no, payment date, payment period, payment delay (since notification or last payment), payment amount, \n",
"\n",
"Time columns:\n",
" * ``occurrence_period`` \n",
" * ``occurrence_time``\n",
" * ``payment_period``\n",
" * ``payment_time``\n",
" * ``setldel``\n",
" * ``notidel``\n",
" \n",
"For the purpose of generality the length of a time unit is unspecified. It could be a month quarter, etc.; one simply adjusts one’s claim frequency to match the choice. The time origin is set at the commencement of the first accident period (which is of unit length); so an occurrence_time of 0.6238 means that the claim occurred 0.6238 of a unit after the commencement of the first accident period. The field ``notidel`` is the number of time units of delay from occurrence to notification. The field ``setldel`` is the number of time units of delay from notification to settlement. The field ``payment_delay`` is the number of time units of delay from notification to the partial payment in the record. The field ``payment_time`` is equal to ``occurrence_time`` + ``notidel`` + ``payment_delay``.\n",
"\n",
"Responses:\n",
" * ``payment_size``: uninflated payments\n",
" * ``payment_inflated``: inflated payments\n",
"\n",
"We could also create a flag for whether the claim is settled."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Transforming data"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"For the models in this article, we will create a dataset with one record per claim number x development period after the claim reported. Such a dataset can be used to predict outstanding case estimates / IBNER.\n",
"\n",
"There are some significant limitations of this approach which we discuss later in the article."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"One key thing to note with this dataset is that **payment delay is measured from notification** and not from date of occurrence. This means that summing up by ``occurence`` and ``payment delay`` will **not** give a standard triangle.\n",
"\n",
"Again, the field ``payment_time`` is equal to ``occurrence_time`` + ``notidel`` + ``payment_delay``.\n",
"\n",
"So set up a ``noti_period`` as ``ceiling(occurrence_time + notidel)`` and define ``development_period`` as ``payment_period - occurrence_period``"
]
},
{
"cell_type": "code",
"execution_count": 111,
"metadata": {},
"outputs": [],
"source": [
"transactions[\"noti_period\"] = np.ceil(transactions[\"occurrence_time\"] + transactions[\"notidel\"]).astype('int')\n",
"transactions[\"settle_period\"] = np.ceil(\n",
" transactions[\"occurrence_time\"] + transactions[\"notidel\"] + transactions[\"setldel\"]\n",
").astype('int')\n",
"transactions[\"development_period\"] = transactions[\"payment_period\"] - transactions[\"occurrence_period\"]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We will use SQL with DuckDB to do part of data wrangling. This is mostly for clarity for this article - SQL code can be easier to read the logic than native pandas code. "
]
},
{
"cell_type": "code",
"execution_count": 80,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<duckdb.DuckDBPyConnection at 0x1678e4970>"
]
},
"execution_count": 80,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# register the table\n",
"con.register('transactions_view', transactions)"
]
},
{
"cell_type": "code",
"execution_count": 112,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"55"
]
},
"execution_count": 112,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"num_dev_periods = (transactions[\"payment_period\"] - transactions[\"occurrence_period\"]).max()\n",
"num_dev_periods"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Group up the claim payments by development period."
]
},
{
"cell_type": "code",
"execution_count": 114,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>development_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>1</td>\n",
" <td>4</td>\n",
" <td>25104.778182</td>\n",
" <td>25631.935128</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>1</td>\n",
" <td>7</td>\n",
" <td>26176.620067</td>\n",
" <td>27112.545886</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>1</td>\n",
" <td>11</td>\n",
" <td>26333.186750</td>\n",
" <td>27828.701791</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>1</td>\n",
" <td>14</td>\n",
" <td>26341.097381</td>\n",
" <td>28293.903794</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>1</td>\n",
" <td>18</td>\n",
" <td>681915.107248</td>\n",
" <td>747368.939460</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16193</th>\n",
" <td>3623</td>\n",
" <td>13</td>\n",
" <td>2191.252889</td>\n",
" <td>47769.049092</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16194</th>\n",
" <td>3624</td>\n",
" <td>1</td>\n",
" <td>8136.387039</td>\n",
" <td>9965.045320</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16195</th>\n",
" <td>3624</td>\n",
" <td>2</td>\n",
" <td>6586.081338</td>\n",
" <td>8093.128975</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16196</th>\n",
" <td>3624</td>\n",
" <td>3</td>\n",
" <td>17487.313820</td>\n",
" <td>21553.216100</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16197</th>\n",
" <td>3624</td>\n",
" <td>4</td>\n",
" <td>238527.509287</td>\n",
" <td>295273.405582</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>16198 rows × 4 columns</p>\n",
"</div>"
],
"text/plain": [
" claim_no development_period payment_size payment_inflated\n",
"0 1 4 25104.778182 25631.935128\n",
"1 1 7 26176.620067 27112.545886\n",
"2 1 11 26333.186750 27828.701791\n",
"3 1 14 26341.097381 28293.903794\n",
"4 1 18 681915.107248 747368.939460\n",
"... ... ... ... ...\n",
"16193 3623 13 2191.252889 47769.049092\n",
"16194 3624 1 8136.387039 9965.045320\n",
"16195 3624 2 6586.081338 8093.128975\n",
"16196 3624 3 17487.313820 21553.216100\n",
"16197 3624 4 238527.509287 295273.405582\n",
"\n",
"[16198 rows x 4 columns]"
]
},
"execution_count": 114,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"\"\"\n",
" CREATE OR REPLACE VIEW transactions_group AS \n",
" SELECT \n",
" claim_no,\n",
" development_period,\n",
" SUM(payment_size) as payment_size, \n",
" SUM(payment_inflated) as payment_inflated\n",
"\n",
" FROM \n",
" transactions\n",
" GROUP BY \n",
" claim_no,\n",
" development_period\n",
" ORDER BY\n",
" claim_no,\n",
" development_period;\n",
" \n",
" SELECT * FROM transactions_group;\n",
"\"\"\"\n",
")\n",
"transactions_group = con.fetchdf()\n",
"transactions_group"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"\"Claims Header\" dataset, but expand development:\n",
" * Include all claims header type information which is fixed for each claim\n",
" * Expand one row per development_period"
]
},
{
"cell_type": "code",
"execution_count": 115,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>settle_period</th>\n",
" <th>occurrence_period</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>2</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>3</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>4</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195197</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>51</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195198</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>52</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195199</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>53</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195200</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>54</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195201</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>55</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>195202 rows × 5 columns</p>\n",
"</div>"
],
"text/plain": [
" claim_no settle_period occurrence_period noti_period \\\n",
"0 1 19 1 1 \n",
"1 1 19 1 1 \n",
"2 1 19 1 1 \n",
"3 1 19 1 1 \n",
"4 1 19 1 1 \n",
"... ... ... ... ... \n",
"195197 3624 44 40 41 \n",
"195198 3624 44 40 41 \n",
"195199 3624 44 40 41 \n",
"195200 3624 44 40 41 \n",
"195201 3624 44 40 41 \n",
"\n",
" development_period \n",
"0 0 \n",
"1 1 \n",
"2 2 \n",
"3 3 \n",
"4 4 \n",
"... ... \n",
"195197 51 \n",
"195198 52 \n",
"195199 53 \n",
"195200 54 \n",
"195201 55 \n",
"\n",
"[195202 rows x 5 columns]"
]
},
"execution_count": 115,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"range_payment_delay = pd.DataFrame.from_dict({\"development_period\": range(0, num_dev_periods + 1)})\n",
"\n",
"con.register('range_payment_delay_view', range_payment_delay)\n",
"\n",
"# create the dataset\n",
"con.execute(\"\"\"\n",
" CREATE OR REPLACE VIEW claim_head_expand_dev AS \n",
" SELECT \n",
" trn.claim_no,\n",
" trn.settle_period, \n",
" trn.occurrence_period,\n",
" trn.noti_period,\n",
" dly.development_period \n",
" FROM \n",
" (\n",
" SELECT DISTINCT \n",
" claim_no, occurrence_period, noti_period, settle_period \n",
" FROM transactions\n",
" ) trn\n",
" JOIN\n",
" range_payment_delay_view dly\n",
" ON \n",
" dly.development_period >= (trn.noti_period - trn.occurrence_period)\n",
" ORDER BY\n",
" trn.claim_no,\n",
" dly.development_period;\n",
" \n",
" SELECT * FROM claim_head_expand_dev;\n",
"\"\"\")\n",
"\n",
"claim_head_expand_dev = con.fetchdf()\n",
"claim_head_expand_dev"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Finally, we get all this together for the expanded table."
]
},
{
"cell_type": "code",
"execution_count": 120,
"metadata": {
"scrolled": false
},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>settle_period</th>\n",
" <th>occurrence_period</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>is_settled</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>2</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>3</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>1</td>\n",
" <td>19</td>\n",
" <td>1</td>\n",
" <td>1</td>\n",
" <td>4</td>\n",
" <td>25104.778182</td>\n",
" <td>25631.935128</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195197</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>51</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195198</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>52</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195199</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>53</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195200</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>54</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>195201</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>55</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>195202 rows × 8 columns</p>\n",
"</div>"
],
"text/plain": [
" claim_no settle_period occurrence_period noti_period \\\n",
"0 1 19 1 1 \n",
"1 1 19 1 1 \n",
"2 1 19 1 1 \n",
"3 1 19 1 1 \n",
"4 1 19 1 1 \n",
"... ... ... ... ... \n",
"195197 3624 44 40 41 \n",
"195198 3624 44 40 41 \n",
"195199 3624 44 40 41 \n",
"195200 3624 44 40 41 \n",
"195201 3624 44 40 41 \n",
"\n",
" development_period payment_size payment_inflated is_settled \n",
"0 0 0.000000 0.000000 0 \n",
"1 1 0.000000 0.000000 0 \n",
"2 2 0.000000 0.000000 0 \n",
"3 3 0.000000 0.000000 0 \n",
"4 4 25104.778182 25631.935128 0 \n",
"... ... ... ... ... \n",
"195197 51 0.000000 0.000000 1 \n",
"195198 52 0.000000 0.000000 1 \n",
"195199 53 0.000000 0.000000 1 \n",
"195200 54 0.000000 0.000000 1 \n",
"195201 55 0.000000 0.000000 1 \n",
"\n",
"[195202 rows x 8 columns]"
]
},
"execution_count": 120,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# create the dataset\n",
"con.execute(\"\"\"\n",
" CREATE OR REPLACE VIEW transactions_expanded AS \n",
" SELECT \n",
" head.*,\n",
" FILL_NULL(trn_grp.payment_size, 0) as payment_size,\n",
" FILL_NULL(trn_grp.payment_inflated, 0) as payment_inflated,\n",
" CASE \n",
" WHEN head.occurrence_period + head.development_period >= head.settle_period THEN 1 \n",
" ELSE 0 \n",
" END as is_settled\n",
"\n",
" FROM \n",
" claim_head_expand_dev as head\n",
" LEFT JOIN \n",
" transactions_group as trn_grp\n",
" ON \n",
" head.claim_no = trn_grp.claim_no\n",
" AND head.development_period = trn_grp.development_period\n",
" ORDER BY \n",
" head.claim_no,\n",
" head.development_period\n",
" ;\n",
" \n",
" SELECT * FROM transactions_expanded;\n",
"\"\"\")\n",
"dat = con.fetchdf()\n",
"dat"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"So what does this do - where does this lead us? \n",
"We basically go from this transactional dataset:"
]
},
{
"cell_type": "code",
"execution_count": 121,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>pmt_no</th>\n",
" <th>occurrence_period</th>\n",
" <th>occurrence_time</th>\n",
" <th>claim_size</th>\n",
" <th>notidel</th>\n",
" <th>setldel</th>\n",
" <th>payment_time</th>\n",
" <th>payment_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>payment_delay</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" <th>settle_period</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>3624</td>\n",
" <td>1</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>40.951591</td>\n",
" <td>41</td>\n",
" <td>8136.387039</td>\n",
" <td>9965.045320</td>\n",
" <td>0.517665</td>\n",
" <td>41</td>\n",
" <td>1</td>\n",
" <td>44</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>3624</td>\n",
" <td>2</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>41.622132</td>\n",
" <td>42</td>\n",
" <td>6586.081338</td>\n",
" <td>8093.128975</td>\n",
" <td>0.670541</td>\n",
" <td>41</td>\n",
" <td>2</td>\n",
" <td>44</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>3624</td>\n",
" <td>3</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>42.081820</td>\n",
" <td>43</td>\n",
" <td>9716.975065</td>\n",
" <td>11967.648057</td>\n",
" <td>0.459688</td>\n",
" <td>41</td>\n",
" <td>3</td>\n",
" <td>44</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>3624</td>\n",
" <td>4</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>42.407479</td>\n",
" <td>43</td>\n",
" <td>7770.338755</td>\n",
" <td>9585.568042</td>\n",
" <td>0.325659</td>\n",
" <td>41</td>\n",
" <td>3</td>\n",
" <td>44</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>3624</td>\n",
" <td>5</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>43.066655</td>\n",
" <td>44</td>\n",
" <td>203618.760893</td>\n",
" <td>252007.199917</td>\n",
" <td>0.659176</td>\n",
" <td>41</td>\n",
" <td>4</td>\n",
" <td>44</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>3624</td>\n",
" <td>6</td>\n",
" <td>40</td>\n",
" <td>39.767468</td>\n",
" <td>270737.291484</td>\n",
" <td>0.666458</td>\n",
" <td>2.920804</td>\n",
" <td>43.354731</td>\n",
" <td>44</td>\n",
" <td>34908.748394</td>\n",
" <td>43266.205665</td>\n",
" <td>0.288075</td>\n",
" <td>41</td>\n",
" <td>4</td>\n",
" <td>44</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" claim_no pmt_no occurrence_period occurrence_time claim_size \\\n",
"0 3624 1 40 39.767468 270737.291484 \n",
"1 3624 2 40 39.767468 270737.291484 \n",
"2 3624 3 40 39.767468 270737.291484 \n",
"3 3624 4 40 39.767468 270737.291484 \n",
"4 3624 5 40 39.767468 270737.291484 \n",
"5 3624 6 40 39.767468 270737.291484 \n",
"\n",
" notidel setldel payment_time payment_period payment_size \\\n",
"0 0.666458 2.920804 40.951591 41 8136.387039 \n",
"1 0.666458 2.920804 41.622132 42 6586.081338 \n",
"2 0.666458 2.920804 42.081820 43 9716.975065 \n",
"3 0.666458 2.920804 42.407479 43 7770.338755 \n",
"4 0.666458 2.920804 43.066655 44 203618.760893 \n",
"5 0.666458 2.920804 43.354731 44 34908.748394 \n",
"\n",
" payment_inflated payment_delay noti_period development_period \\\n",
"0 9965.045320 0.517665 41 1 \n",
"1 8093.128975 0.670541 41 2 \n",
"2 11967.648057 0.459688 41 3 \n",
"3 9585.568042 0.325659 41 3 \n",
"4 252007.199917 0.659176 41 4 \n",
"5 43266.205665 0.288075 41 4 \n",
"\n",
" settle_period \n",
"0 44 \n",
"1 44 \n",
"2 44 \n",
"3 44 \n",
"4 44 \n",
"5 44 "
]
},
"execution_count": 121,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"\"\"SELECT * FROM transactions WHERE claim_no = 3624;\"\"\").fetchdf()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"To one which is grouped by period but also expanded to have dummy rows for all the development periods:"
]
},
{
"cell_type": "code",
"execution_count": 122,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>settle_period</th>\n",
" <th>occurrence_period</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>is_settled</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>1</td>\n",
" <td>8136.387039</td>\n",
" <td>9965.045320</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>2</td>\n",
" <td>6586.081338</td>\n",
" <td>8093.128975</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>3</td>\n",
" <td>17487.313820</td>\n",
" <td>21553.216100</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>4</td>\n",
" <td>238527.509287</td>\n",
" <td>295273.405582</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>5</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>6</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>6</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>7</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>7</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>8</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>9</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>9</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>10</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>10</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>11</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>11</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>12</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>12</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>13</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>13</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>14</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>14</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>15</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>15</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>16</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>17</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>17</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>18</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>19</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>19</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>20</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>20</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>21</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>21</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>22</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>22</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>23</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>23</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>24</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>24</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>25</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>25</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>26</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>26</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>27</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>27</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>28</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>28</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>29</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>29</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>30</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>30</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>31</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>31</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>32</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>32</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>33</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>33</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>34</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>34</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>35</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>35</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>36</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>36</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>37</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>37</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>38</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>38</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>39</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>39</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>40</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>40</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>41</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>41</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>42</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>42</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>43</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>43</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>44</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>44</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>45</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>45</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>46</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>46</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>47</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>47</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>48</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>48</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>49</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>49</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>50</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>50</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>51</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>51</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>52</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>52</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>53</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>53</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>54</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>54</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>55</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" claim_no settle_period occurrence_period noti_period \\\n",
"0 3624 44 40 41 \n",
"1 3624 44 40 41 \n",
"2 3624 44 40 41 \n",
"3 3624 44 40 41 \n",
"4 3624 44 40 41 \n",
"5 3624 44 40 41 \n",
"6 3624 44 40 41 \n",
"7 3624 44 40 41 \n",
"8 3624 44 40 41 \n",
"9 3624 44 40 41 \n",
"10 3624 44 40 41 \n",
"11 3624 44 40 41 \n",
"12 3624 44 40 41 \n",
"13 3624 44 40 41 \n",
"14 3624 44 40 41 \n",
"15 3624 44 40 41 \n",
"16 3624 44 40 41 \n",
"17 3624 44 40 41 \n",
"18 3624 44 40 41 \n",
"19 3624 44 40 41 \n",
"20 3624 44 40 41 \n",
"21 3624 44 40 41 \n",
"22 3624 44 40 41 \n",
"23 3624 44 40 41 \n",
"24 3624 44 40 41 \n",
"25 3624 44 40 41 \n",
"26 3624 44 40 41 \n",
"27 3624 44 40 41 \n",
"28 3624 44 40 41 \n",
"29 3624 44 40 41 \n",
"30 3624 44 40 41 \n",
"31 3624 44 40 41 \n",
"32 3624 44 40 41 \n",
"33 3624 44 40 41 \n",
"34 3624 44 40 41 \n",
"35 3624 44 40 41 \n",
"36 3624 44 40 41 \n",
"37 3624 44 40 41 \n",
"38 3624 44 40 41 \n",
"39 3624 44 40 41 \n",
"40 3624 44 40 41 \n",
"41 3624 44 40 41 \n",
"42 3624 44 40 41 \n",
"43 3624 44 40 41 \n",
"44 3624 44 40 41 \n",
"45 3624 44 40 41 \n",
"46 3624 44 40 41 \n",
"47 3624 44 40 41 \n",
"48 3624 44 40 41 \n",
"49 3624 44 40 41 \n",
"50 3624 44 40 41 \n",
"51 3624 44 40 41 \n",
"52 3624 44 40 41 \n",
"53 3624 44 40 41 \n",
"54 3624 44 40 41 \n",
"\n",
" development_period payment_size payment_inflated is_settled \n",
"0 1 8136.387039 9965.045320 0 \n",
"1 2 6586.081338 8093.128975 0 \n",
"2 3 17487.313820 21553.216100 0 \n",
"3 4 238527.509287 295273.405582 1 \n",
"4 5 0.000000 0.000000 1 \n",
"5 6 0.000000 0.000000 1 \n",
"6 7 0.000000 0.000000 1 \n",
"7 8 0.000000 0.000000 1 \n",
"8 9 0.000000 0.000000 1 \n",
"9 10 0.000000 0.000000 1 \n",
"10 11 0.000000 0.000000 1 \n",
"11 12 0.000000 0.000000 1 \n",
"12 13 0.000000 0.000000 1 \n",
"13 14 0.000000 0.000000 1 \n",
"14 15 0.000000 0.000000 1 \n",
"15 16 0.000000 0.000000 1 \n",
"16 17 0.000000 0.000000 1 \n",
"17 18 0.000000 0.000000 1 \n",
"18 19 0.000000 0.000000 1 \n",
"19 20 0.000000 0.000000 1 \n",
"20 21 0.000000 0.000000 1 \n",
"21 22 0.000000 0.000000 1 \n",
"22 23 0.000000 0.000000 1 \n",
"23 24 0.000000 0.000000 1 \n",
"24 25 0.000000 0.000000 1 \n",
"25 26 0.000000 0.000000 1 \n",
"26 27 0.000000 0.000000 1 \n",
"27 28 0.000000 0.000000 1 \n",
"28 29 0.000000 0.000000 1 \n",
"29 30 0.000000 0.000000 1 \n",
"30 31 0.000000 0.000000 1 \n",
"31 32 0.000000 0.000000 1 \n",
"32 33 0.000000 0.000000 1 \n",
"33 34 0.000000 0.000000 1 \n",
"34 35 0.000000 0.000000 1 \n",
"35 36 0.000000 0.000000 1 \n",
"36 37 0.000000 0.000000 1 \n",
"37 38 0.000000 0.000000 1 \n",
"38 39 0.000000 0.000000 1 \n",
"39 40 0.000000 0.000000 1 \n",
"40 41 0.000000 0.000000 1 \n",
"41 42 0.000000 0.000000 1 \n",
"42 43 0.000000 0.000000 1 \n",
"43 44 0.000000 0.000000 1 \n",
"44 45 0.000000 0.000000 1 \n",
"45 46 0.000000 0.000000 1 \n",
"46 47 0.000000 0.000000 1 \n",
"47 48 0.000000 0.000000 1 \n",
"48 49 0.000000 0.000000 1 \n",
"49 50 0.000000 0.000000 1 \n",
"50 51 0.000000 0.000000 1 \n",
"51 52 0.000000 0.000000 1 \n",
"52 53 0.000000 0.000000 1 \n",
"53 54 0.000000 0.000000 1 \n",
"54 55 0.000000 0.000000 1 "
]
},
"execution_count": 122,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"\"\"SELECT * FROM transactions_expanded WHERE claim_no = 3624;\"\"\").fetchdf()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"So when applied in practice the model might see a truncated version of the above, like this:"
]
},
{
"cell_type": "code",
"execution_count": 139,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>settle_period</th>\n",
" <th>occurrence_period</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>is_settled</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>1</td>\n",
" <td>8136.387039</td>\n",
" <td>9965.045320</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>2</td>\n",
" <td>6586.081338</td>\n",
" <td>8093.128975</td>\n",
" <td>0</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" claim_no settle_period occurrence_period noti_period \\\n",
"0 3624 44 40 41 \n",
"1 3624 44 40 41 \n",
"\n",
" development_period payment_size payment_inflated is_settled \n",
"0 1 8136.387039 9965.045320 0 \n",
"1 2 6586.081338 8093.128975 0 "
]
},
"execution_count": 139,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"SELECT * FROM transactions_expanded WHERE claim_no = 3624 and development_period <= 2;\").fetchdf()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"And it would try to predict this:"
]
},
{
"cell_type": "code",
"execution_count": 140,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>claim_no</th>\n",
" <th>settle_period</th>\n",
" <th>occurrence_period</th>\n",
" <th>noti_period</th>\n",
" <th>development_period</th>\n",
" <th>payment_size</th>\n",
" <th>payment_inflated</th>\n",
" <th>is_settled</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>3</td>\n",
" <td>17487.313820</td>\n",
" <td>21553.216100</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>4</td>\n",
" <td>238527.509287</td>\n",
" <td>295273.405582</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>5</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>6</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>7</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>8</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>6</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>9</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>7</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>10</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>11</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>9</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>12</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>10</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>13</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>11</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>14</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>12</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>15</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>13</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>16</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>14</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>17</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>15</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>18</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>19</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>17</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>20</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>21</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>19</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>22</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>20</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>23</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>21</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>24</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>22</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>25</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>23</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>26</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>24</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>27</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>25</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>28</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>26</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>29</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>27</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>30</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>28</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>31</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>29</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>32</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>30</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>33</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>31</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>34</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>32</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>35</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>33</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>36</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>34</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>37</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>35</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>38</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>36</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>39</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>37</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>40</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>38</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>41</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>39</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>42</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>40</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>43</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>41</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>44</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>42</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>45</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>43</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>46</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>44</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>47</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>45</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>48</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>46</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>49</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>47</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>50</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>48</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>51</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>49</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>52</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>50</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>53</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>51</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>54</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>52</th>\n",
" <td>3624</td>\n",
" <td>44</td>\n",
" <td>40</td>\n",
" <td>41</td>\n",
" <td>55</td>\n",
" <td>0.000000</td>\n",
" <td>0.000000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" claim_no settle_period occurrence_period noti_period \\\n",
"0 3624 44 40 41 \n",
"1 3624 44 40 41 \n",
"2 3624 44 40 41 \n",
"3 3624 44 40 41 \n",
"4 3624 44 40 41 \n",
"5 3624 44 40 41 \n",
"6 3624 44 40 41 \n",
"7 3624 44 40 41 \n",
"8 3624 44 40 41 \n",
"9 3624 44 40 41 \n",
"10 3624 44 40 41 \n",
"11 3624 44 40 41 \n",
"12 3624 44 40 41 \n",
"13 3624 44 40 41 \n",
"14 3624 44 40 41 \n",
"15 3624 44 40 41 \n",
"16 3624 44 40 41 \n",
"17 3624 44 40 41 \n",
"18 3624 44 40 41 \n",
"19 3624 44 40 41 \n",
"20 3624 44 40 41 \n",
"21 3624 44 40 41 \n",
"22 3624 44 40 41 \n",
"23 3624 44 40 41 \n",
"24 3624 44 40 41 \n",
"25 3624 44 40 41 \n",
"26 3624 44 40 41 \n",
"27 3624 44 40 41 \n",
"28 3624 44 40 41 \n",
"29 3624 44 40 41 \n",
"30 3624 44 40 41 \n",
"31 3624 44 40 41 \n",
"32 3624 44 40 41 \n",
"33 3624 44 40 41 \n",
"34 3624 44 40 41 \n",
"35 3624 44 40 41 \n",
"36 3624 44 40 41 \n",
"37 3624 44 40 41 \n",
"38 3624 44 40 41 \n",
"39 3624 44 40 41 \n",
"40 3624 44 40 41 \n",
"41 3624 44 40 41 \n",
"42 3624 44 40 41 \n",
"43 3624 44 40 41 \n",
"44 3624 44 40 41 \n",
"45 3624 44 40 41 \n",
"46 3624 44 40 41 \n",
"47 3624 44 40 41 \n",
"48 3624 44 40 41 \n",
"49 3624 44 40 41 \n",
"50 3624 44 40 41 \n",
"51 3624 44 40 41 \n",
"52 3624 44 40 41 \n",
"\n",
" development_period payment_size payment_inflated is_settled \n",
"0 3 17487.313820 21553.216100 0 \n",
"1 4 238527.509287 295273.405582 1 \n",
"2 5 0.000000 0.000000 1 \n",
"3 6 0.000000 0.000000 1 \n",
"4 7 0.000000 0.000000 1 \n",
"5 8 0.000000 0.000000 1 \n",
"6 9 0.000000 0.000000 1 \n",
"7 10 0.000000 0.000000 1 \n",
"8 11 0.000000 0.000000 1 \n",
"9 12 0.000000 0.000000 1 \n",
"10 13 0.000000 0.000000 1 \n",
"11 14 0.000000 0.000000 1 \n",
"12 15 0.000000 0.000000 1 \n",
"13 16 0.000000 0.000000 1 \n",
"14 17 0.000000 0.000000 1 \n",
"15 18 0.000000 0.000000 1 \n",
"16 19 0.000000 0.000000 1 \n",
"17 20 0.000000 0.000000 1 \n",
"18 21 0.000000 0.000000 1 \n",
"19 22 0.000000 0.000000 1 \n",
"20 23 0.000000 0.000000 1 \n",
"21 24 0.000000 0.000000 1 \n",
"22 25 0.000000 0.000000 1 \n",
"23 26 0.000000 0.000000 1 \n",
"24 27 0.000000 0.000000 1 \n",
"25 28 0.000000 0.000000 1 \n",
"26 29 0.000000 0.000000 1 \n",
"27 30 0.000000 0.000000 1 \n",
"28 31 0.000000 0.000000 1 \n",
"29 32 0.000000 0.000000 1 \n",
"30 33 0.000000 0.000000 1 \n",
"31 34 0.000000 0.000000 1 \n",
"32 35 0.000000 0.000000 1 \n",
"33 36 0.000000 0.000000 1 \n",
"34 37 0.000000 0.000000 1 \n",
"35 38 0.000000 0.000000 1 \n",
"36 39 0.000000 0.000000 1 \n",
"37 40 0.000000 0.000000 1 \n",
"38 41 0.000000 0.000000 1 \n",
"39 42 0.000000 0.000000 1 \n",
"40 43 0.000000 0.000000 1 \n",
"41 44 0.000000 0.000000 1 \n",
"42 45 0.000000 0.000000 1 \n",
"43 46 0.000000 0.000000 1 \n",
"44 47 0.000000 0.000000 1 \n",
"45 48 0.000000 0.000000 1 \n",
"46 49 0.000000 0.000000 1 \n",
"47 50 0.000000 0.000000 1 \n",
"48 51 0.000000 0.000000 1 \n",
"49 52 0.000000 0.000000 1 \n",
"50 53 0.000000 0.000000 1 \n",
"51 54 0.000000 0.000000 1 \n",
"52 55 0.000000 0.000000 1 "
]
},
"execution_count": 140,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"SELECT * FROM transactions_expanded WHERE claim_no = 3624 and development_period > 2;\").fetchdf()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"With this structure, the models will predict IBNER only."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Quick check - do I have all my claims?"
]
},
{
"cell_type": "code",
"execution_count": 124,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>sum(payment_size)</th>\n",
" <th>sum(payment_inflated)</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>5.745091e+08</td>\n",
" <td>1.091564e+09</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" sum(payment_size) sum(payment_inflated)\n",
"0 5.745091e+08 1.091564e+09"
]
},
"execution_count": 124,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"SELECT SUM(payment_size), SUM(payment_inflated) FROM transactions;\").fetchdf()"
]
},
{
"cell_type": "code",
"execution_count": 125,
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>sum(payment_size)</th>\n",
" <th>sum(payment_inflated)</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>5.745091e+08</td>\n",
" <td>1.091564e+09</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" sum(payment_size) sum(payment_inflated)\n",
"0 5.745091e+08 1.091564e+09"
]
},
"execution_count": 125,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"con.execute(\"SELECT SUM(payment_size), SUM(payment_inflated) FROM transactions_expanded;\").fetchdf()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The totals appear to match up."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Limitations of this data structure\n",
"\n",
"The nice thing about this relatively simple tabular design is that it does sum up nicely as we will see in the follow up plots. Compared to the triangles used in earlier articles, we have individual claims data.\n",
"\n",
"Several limitations become apparent however. We only have ``occurrence_period``, ``noti_period``, and ``development_period`` as features. We do not have claim descriptions or other categorisation codes for the claim.\n",
"\n",
"Within the tabular structure to follow, we do not have the transaction history as features - this is a significant limitation. Without the open/closed flag as a feature, the model will predict the same IBNER for closed claims as it does for open claims and the IBNER predictions will only make sense at the aggregate level. \n",
"\n",
"To utilise transaction history, we would need to try a different tabular formulation of the problem. An example of an alternative tabular formation that could use transaction history would be to have:\n",
"\n",
" * Snapshots of the claim, \n",
" * Record per claim transaction with cumulative paid and settlement status at each time\n",
" * Remaining claims cost of that claim as response\n",
" \n",
"Another alternative would be to explore RNN-style structures. "
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Train / test splits"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"There are a number of ways the data can be split for training vs testing. \n",
"\n",
"Here we will split it by claim number. Entire claims are placed into either the train or the test set. This means the holdout data is entirely unseen claims.\n",
"\n",
"However, having the entire claims history for training data claims provides models witih more of the tail as training data than they may otherwise in a practical context.\n",
"\n",
"Another approach to the test set may be a cut-off by calendar period - to mask the bottom half of the triangle."
]
},
{
"cell_type": "code",
"execution_count": 143,
"metadata": {},
"outputs": [],
"source": [
"# Set up train and test indicator\n",
"# Split by whole claim\n",
"dat[\"train_ind\"] = (dat.claim_no % 10 >= 3)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Model datasets"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"We create the model datasets here."
]
},
{
"cell_type": "code",
"execution_count": 135,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"55"
]
},
"execution_count": 135,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Number of acc/dev periods\n",
"num_occ_periods = int(dat[\"occurrence_period\"].max()) \n",
"num_dev_periods = int(dat[\"development_period\"].max())\n",
"\n",
"num_periods = max([num_occ_periods, num_dev_periods])\n",
"num_periods"
]
},
{
"cell_type": "code",
"execution_count": 136,
"metadata": {},
"outputs": [],
"source": [
"list_of_features = [\"occurrence_period\", \"development_period\", \"noti_period\"]"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"In terms of outputs we have ``payment_size``, ``payment_inflated`` or ``is_settled`` as potential candidates.\n",
"\n",
"One notable feature of neural networks is the ability to predict multiple outputs. See [this article for an example](https://actuariesinstitute.github.io/cookbook/docs/multitasking_risk_pricing.html). In this article we will try several different model structures, but for simplicity only predict the uninflated payments in this notebook. "
]
},
{
"cell_type": "code",
"execution_count": 144,
"metadata": {
"id": "vh0YfVxJjGma"
},
"outputs": [],
"source": [
"# Set up train and test sets\n",
"X_train = (dat.loc[dat.train_ind, list_of_features])\n",
"y_train = (dat.loc[dat.train_ind, \"payment_size\"])\n",
"\n",
"X_test = (dat.loc[dat.train_ind == False, list_of_features])\n",
"y_test = (dat.loc[dat.train_ind == False, \"payment_size\"])\n",
"\n",
"X = (dat.loc[:, list_of_features])\n",
"y = (dat.loc[:, \"payment_size\"])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The dataset looks like this on the log-scale:"
]
},
{
"cell_type": "code",
"execution_count": 160,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x169ad26d0>"
]
},
"execution_count": 160,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"(\n",
" dat\n",
" .groupby([\"occurrence_period\", \"development_period\"])\n",
" .agg('sum')\n",
" .reset_index()\n",
" .pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"payment_size\")\n",
" .plot(logy=True)\n",
")\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Compared to the dataset used in the previous LASSO article, this dataset has considerably higher claims volatility, but the LASSO dataset has the unusual spike in claims."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "wcMCrLQl-taP"
},
"source": [
"## Tabular deep learning with scikit-learn"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "zMJbVd7y9toE"
},
"source": [
"This article focuses on tabular neural networks. Like ``scikit-learn`` these are models that predict a ``y`` from an ``X`` with fixed dimensions. There are other kinds of neural networks - for example a [Recurrent Neural Network](https://www.ibm.com/cloud/learn/recurrent-neural-networks) is analogous to an ARIMA and models timeseries or sequence data.\n",
"\n",
"So we will again, make a **scikit-learn** Regressor with a modified version of scikit learn wrapper and plug in the neural networks within the framework. \n",
"\n",
"The key components are:\n",
"\n",
" * ``__init__``: Define any hyperparameters here. At a minimum, this would include the lasso regularization penalty.\n",
"\n",
" * ``fit``: Define the training process for the model. We will put the actual logic in partial fit (and try a few times as sometimes neural networks are finicky and blow up).\n",
"\n",
" * ``partial_fit``: For ``pytorch`` we have a training loop where we read data, calculated the loss, backpropogate it and update the weights.\n",
"\n",
" * ``predict``: Define how to get predictions - i.e. apply the ``forward`` method of the Module.\n",
"\n",
" * ``score``: Define performance. Here we will use RMSE for consistency with the notebook.\n",
"\n",
"There is also added boilerplate to get data into the right formats."
]
},
{
"cell_type": "code",
"execution_count": 207,
"metadata": {
"id": "vw4mWOZQ9anS"
},
"outputs": [],
"source": [
"class TabularNetRegressor(BaseEstimator, RegressorMixin):\n",
"\n",
" def __init__(\n",
" self, \n",
" module,\n",
" criterion=nn.PoissonNLLLoss,\n",
" l1_penalty=0.0, # lambda is a reserved word\n",
" l1_applies_params=[\"linear.weight\", \"neural.weight\", \"neural1.weight\", \"neural2.weight\"],\n",
" weight_decay=0.0,\n",
" n_hidden=250,\n",
" \n",
" max_iter=100, \n",
" max_lr=0.01,\n",
" \n",
" attempts=3,\n",
" \n",
" verbose=1,\n",
" print_loss_every_iter=10, \n",
" target_device=torch.device(\"cuda\") if torch.cuda.is_available() else torch.device(\"cpu\"), # Use GPU if available\n",
" **kwargs\n",
" ):\n",
" \"\"\" Tabular Neural Network Regressor (for Claims Reserving)\n",
"\n",
" This trains a neural network with specified loss, Log Link and l1 LASSO penalties\n",
" using Pytorch. It has early stopping.\n",
"\n",
" Args:\n",
" module: pytorch nn.Module. Should have n_input as a parameter and\n",
" if l1_penalty, init_weight, or init_bias are used, a final layer \n",
" called \"linear\".\n",
"\n",
" criterion: pytorch loss function. Consider nn.PoissonNLLLoss for log link.\n",
"\n",
" l1_penalty (float): l1 penalty factor. If not zero, is applied to \n",
" the layers in the Module with names matching l1_applies_params.\n",
"\n",
" (we use l1_penalty because lambda is a reserved word in Python \n",
" for anonymous functions)\n",
"\n",
" weight_decay (float): weight decay - analogous to l2 penalty factor\n",
" Applied to all weights\n",
" \n",
" max_iter (int): Maximum number of epochs before training stops. \n",
" Previously this used a high value for triangles since the record count is so small.\n",
" For larger regression problems, a lower number of iterations may be sufficient.\n",
"\n",
" n_hidden (int): Passed to module. Hidden layer size.\n",
"\n",
" max_lr (float): Min / Max learning rate - we will use one_cycle_lr\n",
"\n",
" attempts (int): Try fitting the model this many times before giving up and throwing error\n",
"\n",
" verbose (int): 0 means don't print. 1 means do print.\n",
" \n",
" print_loss_every_iter (int): Print the loss very x epochs\n",
"\n",
" \"\"\"\n",
" self.module = module\n",
" self.criterion = criterion\n",
" self.l1_penalty = l1_penalty\n",
" self.l1_applies_params = l1_applies_params\n",
" self.weight_decay = weight_decay\n",
" self.max_iter = max_iter\n",
" self.target_device = target_device\n",
" self.n_hidden = n_hidden \n",
" self.max_lr = max_lr\n",
" self.attempts = attempts\n",
" self.print_loss_every_iter = print_loss_every_iter\n",
" self.verbose = verbose\n",
" self.kwargs = kwargs\n",
"\n",
" \n",
" def fix_array(self, y):\n",
" \"Need to be picky about array formats\"\n",
" if isinstance(y, pd.DataFrame) or isinstance(y, pd.Series):\n",
" y = y.values\n",
" if y.ndim == 1:\n",
" y = y.reshape(-1, 1)\n",
" y = y.astype(np.float32)\n",
" return y\n",
" \n",
"\n",
" def fit(self, X, y):\n",
" # The main fit logic is in partial_fit\n",
" # We will try a few times if numbers explode because NN's are finicky and we are doing CV\n",
" n_input = X.shape[-1]\n",
"\n",
" for attempt in range(0, self.attempts):\n",
" try:\n",
" # Training new model\n",
" self.module_ = self.module(\n",
" n_input=n_input, \n",
" n_hidden=self.n_hidden,\n",
" init_bias=np.log(y.mean()),\n",
" **self.kwargs\n",
" ).to(self.target_device)\n",
"\n",
" # Partial fit means you take an existing model and keep training \n",
" # so the logic is basically the same\n",
" self.partial_fit(X, y)\n",
"\n",
" break\n",
" except ValueError:\n",
" if attempt < self.attempts - 1:\n",
" pass\n",
" if self.verbose > 0: \n",
" print(\"nan loss - trying again\")\n",
" else:\n",
" raise ValueError('Error: nan loss')\n",
"\n",
" return self\n",
"\n",
"\n",
" def partial_fit(self, X, y):\n",
"\n",
" # Check that X and y have correct shape\n",
" X, y = check_X_y(X, y)\n",
"\n",
" # Convert to Pytorch Tensor\n",
" X_tensor = torch.from_numpy(self.fix_array(X)).to(self.target_device)\n",
" y_tensor = torch.from_numpy(self.fix_array(y)).to(self.target_device)\n",
"\n",
" # Optimizer - the generically useful AdamW. Other options like SGD\n",
" # are also possible.\n",
" optimizer = torch.optim.AdamW(\n",
" params=self.module_.parameters(),\n",
" lr=self.max_lr / 10,\n",
" weight_decay=self.weight_decay\n",
" )\n",
" \n",
" # Scheduler - one cycle LR\n",
" scheduler = torch.optim.lr_scheduler.OneCycleLR(\n",
" optimizer, \n",
" max_lr=self.max_lr, \n",
" steps_per_epoch=1, \n",
" epochs=self.max_iter\n",
" )\n",
"\n",
" # Loss Function\n",
" try:\n",
" loss_fn = self.criterion(log_input=False).to(self.target_device) # Pytorch loss function\n",
" except TypeError:\n",
" loss_fn = self.criterion # Custom loss function\n",
"\n",
"\n",
" # Training loop\n",
" for epoch in range(self.max_iter): # Repeat max_iter times\n",
" self.module_.train()\n",
" y_pred = self.module_(X_tensor) # Apply current model\n",
" loss = loss_fn(y_pred, y_tensor) # What is the loss on it?\n",
" if self.l1_penalty > 0.0: # Lasso penalty\n",
" loss += self.l1_penalty * sum(\n",
" [\n",
" w.abs().sum()\n",
" for p, w in self.module_.named_parameters()\n",
" if p in self.l1_applies_params\n",
" ]\n",
" )\n",
" \n",
" optimizer.zero_grad() # Reset optimizer\n",
" loss.backward() # Apply back propagation\n",
" optimizer.step() # Update model parameters\n",
" scheduler.step()\n",
"\n",
" if torch.isnan(loss.data).tolist():\n",
" raise ValueError('Error: nan loss')\n",
"\n",
" self.module_.eval() # Eval mode \n",
" y_pred_point = self.module_(X_tensor) # Get \"real\" model estimates\n",
" rmse = torch.sqrt(torch.mean(torch.square(y_pred_point - y_tensor)))\n",
" self.module_.train() # back to distribution\n",
" \n",
" # Every self.print_loss_every_iter steps, print RMSE \n",
" if (epoch % self.print_loss_every_iter == 0) and (self.verbose > 0):\n",
" print(\"Train RMSE: \", rmse.data.tolist(), \" Train Loss: \", loss.data.tolist())\n",
" \n",
" # Return the regressor\n",
" return self\n",
"\n",
"\n",
" def predict(self, X):\n",
" # Checks\n",
" check_is_fitted(self) # Check is fit had been called\n",
" X = check_array(X) # Check input\n",
"\n",
" # Convert to Pytorch Tensor\n",
" X_tensor = torch.from_numpy(self.fix_array(X)).to(self.target_device)\n",
" \n",
" self.module_.eval() # Eval (prediction) mode\n",
"\n",
" # Apply current model and convert back to numpy\n",
" y_pred = self.module_(X_tensor).cpu().detach().numpy().ravel()\n",
" \n",
" return y_pred\n",
"\n",
"\n",
" def score(self, X, y):\n",
" # Negative RMSE score (higher needs to be better)\n",
" y_pred = self.predict(X)\n",
" return -np.sqrt(np.mean((y_pred - y)**2))"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "Fl9DzKW5q0VB"
},
"source": [
"## Lasso model with ramps\n",
"\n",
"In the previous article, we created \"ramp\" splines and heaviside interaction functions for the LASSO model, which performed well. The code is included again below:"
]
},
{
"cell_type": "code",
"execution_count": 208,
"metadata": {
"id": "Tn6LSkb38h8j"
},
"outputs": [],
"source": [
"def LinearSpline(var, start, stop):\n",
" \"\"\"\n",
" Linear spline function - used in data generation and in spline generation below\n",
" \"\"\"\n",
" return np.minimum(stop - start, np.maximum(0, var - start))\n",
"\n",
"\n",
"def GetScaling(vec):\n",
" \"\"\"\n",
" Function to calculate scaling factors for the basis functions\n",
" scaling is discussed in the paper\n",
" \"\"\"\n",
" fn = len(vec)\n",
" fm = np.mean(vec)\n",
" fc = vec - fm \n",
"\n",
" return ((np.sum(fc**2))/fn)**0.5\n",
"\n",
"\n",
"def GetRamps(vec, vecname, nperiods, scaling):\n",
" \"\"\"\n",
" Function to create the ramps for a particular primary vector\n",
" vec = fundamental regressor\n",
" vecname = name of regressor\n",
" np = number of periods\n",
" scaling = scaling factor to use\n",
" \"\"\"\n",
" df = pd.DataFrame.from_dict(\n",
" {\n",
" f\"L_{i}_999_{vecname}\": LinearSpline(vec, i, 999) / scaling\n",
" for i in range(1, nperiods)\n",
" }\n",
" )\n",
" return df\n",
"\n",
"\n",
"def GetInts(vec1, vec2, vecname1, vecname2, nperiods, scaling1, scaling2, train_ind):\n",
" \"\"\"\n",
" Create the step (heaviside) function interactions.\n",
" f\"I_{vecname1}_ge_{i}xI_{vecname2}_ge_{j}\" formats the name of the column\n",
" LinearSpline(vec1, 1, i+1) / scaling1 * LinearSpline(vec2, j, j + 1) / scaling2\n",
" is the interaction term\n",
" and we loop over all combinations of 1:nperiods and 2:nperiods\n",
" \"\"\"\n",
" \n",
" vecs = {}\n",
" for i, j in itertools.product(*[range(2, nperiods), range(2, nperiods)]):\n",
" interaction = (\n",
" LinearSpline(vec1, i - 1, i) / scaling1 * \n",
" LinearSpline(vec2, j - 1, j) / scaling2\n",
" )\n",
"\n",
" # Only include if non-constant over training data\n",
" if not np.all(interaction[train_ind] == interaction[0]):\n",
" vecs[f\"I_{vecname1}_ge_{i}xI_{vecname2}_ge_{j}\"] = interaction\n",
"\n",
" df = pd.DataFrame.from_dict(vecs)\n",
" return df"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"In this article we will make these more scikit-learn friendly. For use in pipelines, or other models, these ramp functions can be adapted to create a ``scikit-learn`` **Transformer**. Similar to setting up a custom **Regressor**,\n",
" * ``__init__`` sets the \"hyperparameters\"\n",
" * ``fit`` sets the parameters\n",
" * ``transform`` applies the parameters on new data"
]
},
{
"cell_type": "code",
"execution_count": 209,
"metadata": {
"id": "geH6P0aA9naM"
},
"outputs": [],
"source": [
"class RampTransformer(BaseEstimator, TransformerMixin):\n",
" def __init__(self, num_periods):\n",
" self.num_periods = num_periods\n",
"\n",
" def fit(self, X, y=None):\n",
" # stash training set\n",
" self.dat = pd.DataFrame(X) \n",
" \n",
" # get the scaling values\n",
" self.rho_factors = {\n",
" v: GetScaling(dat[v].values) for v in dat.columns\n",
" }\n",
" return self\n",
"\n",
" def transform(self, X, y=None):\n",
" dat = pd.concat(\n",
" [\n",
" pd.DataFrame(X).assign(__train_ind=False),\n",
" self.dat.assign(__train_ind=True)\n",
" ], \n",
" axis=\"rows\").reset_index()\n",
" \n",
" # main effects - matrix of values\n",
" main_effects = pd.concat([\n",
" GetRamps(\n",
" vec=dat[v], \n",
" vecname=v, \n",
" nperiods=self.num_periods, \n",
" scaling=self.rho_factors[v]\n",
" ) for v in self.dat.columns\n",
" ], axis=\"columns\")\n",
"\n",
" # interaction effects\n",
" int_effects = pd.concat([\n",
" GetInts(\n",
" vec1=dat[v1], vecname1=v1, scaling1=self.rho_factors[v1], \n",
" vec2=dat[v2], vecname2=v2, scaling2=self.rho_factors[v2], \n",
" nperiods=self.num_periods, train_ind=dat[\"__train_ind\"]\n",
" ) for v1, v2 in itertools.combinations(self.dat.columns, 2)\n",
" ], axis=\"columns\")\n",
"\n",
" varset = pd.concat([main_effects, int_effects], axis=\"columns\")\n",
"\n",
" return varset.loc[dat[\"__train_ind\"]==False]\n"
]
},
{
"cell_type": "code",
"execution_count": 210,
"metadata": {
"id": "Tu7qKuAEP6-0"
},
"outputs": [],
"source": [
"class ColumnKeeper(BaseEstimator, TransformerMixin):\n",
" \"\"\"\n",
" Keeps named cols, preserves DataFrame output\n",
" \"\"\"\n",
" def __init__(self, cols):\n",
" self.cols = cols\n",
"\n",
" def fit(self, X, y):\n",
" return self\n",
"\n",
" def transform(self, X):\n",
" X = X.copy()\n",
" return X[self.cols]"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "XpkZT7zoXWj8"
},
"source": [
"In the next section we will finally set up some neural network models, and see how they fare."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "gdnny3doz90U"
},
"source": [
"## Networks"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "PUxxBefL9n7F"
},
"source": [
"### Log Link Generalized Linear Model"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "JR023rZeM3qU"
},
"source": [
"Neural networks are Pytorch \"Modules\". The key features with these are:\n",
" * Each module is a Python class. \n",
" * The initialisation ``__init__`` stores hyperparameters and defines the parameters of the model. \n",
" * The ``forward`` step defines the \"forward pass\" of the model.\n",
"\n",
"We start off with a simple module that is not even a neural network - but instead a Log Link GLM, which can be used for a LASSO model. With the ramp and interaction transformations to the data, this model performed the best with the dataset in the previous article."
]
},
{
"cell_type": "code",
"execution_count": 211,
"metadata": {
"id": "ict1Otjz8nYH"
},
"outputs": [],
"source": [
"class LogLinkGLM(nn.Module):\n",
" # Define the parameters in __init__\n",
" def __init__(\n",
" self, \n",
" n_input=3, # number of inputs\n",
" n_output=1, # number of outputs\n",
" init_bias=0.0, # init mean value to speed up convergence\n",
" **kwargs, # Ignored; Not used\n",
" ):\n",
"\n",
" super(LogLinkGLM, self).__init__()\n",
" \n",
" self.linear = torch.nn.Linear(n_input, n_output) # Linear coefficients\n",
" nn.init.zeros_(self.linear.weight) # Initialise to zero\n",
" nn.init.constant_(self.linear.bias, init_bias) \n",
"\n",
"\n",
" # The forward functions defines how you get y from X.\n",
" def forward(self, x):\n",
" return torch.exp(self.linear(x)) # log(Y) = XB -> Y = exp(XB)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "S9pVFdq40HeP"
},
"source": [
"### Log Link Feedforward Neural Network"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "GSuqXSiS-fDK"
},
"source": [
"In the last article, we had a basic feedforward model using ``scikit-learn``'s ``MLPRegressor``. \n",
"\n",
"Here, we use Pytorch with some enhancements. From an input ``x``, the model will:\n",
"\n",
" 1. Apply a linear transformation function\n",
" 2. Apply an activation function which provides the non-linearity of the neural network. The ReLU function ``= max(0, z)`` is a common choice.\n",
" 3. Apply a **batch-normalisation** layer to normalise the output\n",
" 4. Apply a **log-link** GLM to the neural-network transformed factors for final output.\n",
"\n",
"Feedforward models can have multiple hidden layers (repeat steps 1-4) which can help with modelling higher complexity relationships, but first we will have some models with only one hidden layer."
]
},
{
"cell_type": "code",
"execution_count": 212,
"metadata": {
"id": "jcPGwkPBROrL"
},
"outputs": [],
"source": [
"class LogLinkForwardNet(nn.Module):\n",
" # Define the parameters in __init__\n",
" def __init__(\n",
" self, \n",
" n_hidden, # hidden layer size\n",
" n_input=3, # number of inputs\n",
" n_output=1, # number of outputs\n",
" init_bias=0.0, # init mean value to speed up convergence \n",
" ): \n",
"\n",
" super(LogLinkForwardNet, self).__init__()\n",
"\n",
" self.hidden = torch.nn.Linear(n_input, n_hidden) # Hidden layer\n",
" self.batchn = torch.nn.BatchNorm1d(n_hidden) # Batchnorm layer\n",
"\n",
" self.linear = torch.nn.Linear(n_hidden, n_output) # Linear coefficients\n",
"\n",
" nn.init.zeros_(self.linear.weight) # Initialise to zero\n",
" nn.init.constant_(self.linear.bias, init_bias) \n",
"\n",
" # The forward function defines how you get y from X.\n",
" def forward(self, x):\n",
" h = F.relu(self.hidden(x)) # Apply hidden layer \n",
" h = self.batchn(h) # apply batchnorm\n",
" \n",
" return torch.exp(self.linear(h)) # log(Y) = XB -> Y = exp(XB)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "FwpOv1Fh0Oyq"
},
"source": [
"### Log Link Residual Network (ResNet)\n",
"\n",
"A [Residual Neural Network, or \"ResNet\"](https://towardsdatascience.com/understanding-and-visualizing-resnets-442284831be8) has a *skip connection* or *shortcut* to the network, which allows data to jump over the hidden layer.\n",
"\n",
"Without the hidden layer, the model works like a GLM. With the hidden layer, it is a feedforward neural network. So ResNet works much like a GLM and a neural network added together."
]
},
{
"cell_type": "code",
"execution_count": 213,
"metadata": {
"id": "nYIvVOhF0U7F"
},
"outputs": [],
"source": [
"class LogLinkResNet(nn.Module):\n",
" # Define the parameters in __init__\n",
" def __init__(\n",
" self, \n",
" n_hidden, # hidden layer size \n",
" n_input=3, # number of inputs\n",
" n_output=1, # number of outputs\n",
" init_bias=0.0, # init mean value to speed up convergence \n",
" ): \n",
"\n",
" super(LogLinkResNet, self).__init__()\n",
"\n",
" self.hidden = torch.nn.Linear(n_input, n_hidden) # Hidden layer\n",
" self.batchn = torch.nn.BatchNorm1d(n_hidden) # Batchnorm layer \n",
" self.linear = torch.nn.Linear(n_input, n_output) # Linear coefficients\n",
"\n",
" # Neural net coefficients - no bias - glm has a bias already\n",
" self.neural = torch.nn.Linear(n_hidden, n_output, bias=False) \n",
"\n",
" nn.init.zeros_(self.linear.weight) # Initialise to zero\n",
" nn.init.zeros_(self.neural.weight) # Initialise to zero\n",
" # n.b. do not initialise hidden layers to zero\n",
" nn.init.constant_(self.linear.bias, init_bias) \n",
"\n",
"\n",
" # The forward function defines how you get y from X.\n",
" def forward(self, x):\n",
" h = F.relu(self.hidden(x)) # Apply hidden layer \n",
" h = self.batchn(h) # Apply batchnorm \n",
" return torch.exp(self.linear(x) + self.neural(h)) # Add GLM to NN"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "L7kY86OIVR9W"
},
"source": [
"### Log Link DenseNet\n",
"\n",
"The DenseNet is a similar idea to ResNets, except outputs are concatenated instead of added. \n",
"\n",
"This is like including NN outputs as GLM features, together with the original features."
]
},
{
"cell_type": "code",
"execution_count": 214,
"metadata": {
"id": "NhTz1IWoVRir"
},
"outputs": [],
"source": [
"class LogLinkDenseNet(nn.Module):\n",
" # Define the parameters in __init__\n",
" def __init__(\n",
" self, \n",
" n_hidden, # hidden layer size\n",
" n_input=3, # number of inputs\n",
" n_output=1, # number of outputs\n",
" init_bias=0.0, # init mean value to speed up convergence \n",
" ): \n",
"\n",
" super(LogLinkDenseNet, self).__init__()\n",
"\n",
" self.hidden = torch.nn.Linear(n_input, n_hidden) # Hidden layer\n",
" self.batchn = torch.nn.BatchNorm1d(n_hidden) # Batchnorm layer \n",
" self.linear = torch.nn.Linear(n_hidden + n_input, n_output) # Expands input\n",
"\n",
" nn.init.zeros_(self.linear.weight) # Initialise to zero\n",
" nn.init.constant_(self.linear.bias, init_bias)\n",
"\n",
" # The forward function defines how you get y from X.\n",
" def forward(self, x):\n",
" h = F.relu(self.hidden(x)) # Apply hidden layer \n",
" h = self.batchn(h) # Apply batchnorm \n",
" h = self.linear(torch.cat((x, h), dim=1)) # Include NN hidden layer with original GLM features\n",
" return torch.exp(h) # Output"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "HxvV8Hz-8XN8"
},
"source": [
"### More Layers\n",
"\n",
"Expanding the ResNet logic to two hidden layers would typically look something like the below. Adding extra layers can help model complex effects, but can overfit (especially with the small size of the dataset). For the second hidden layer we will introduce a sigmoid activation to try to capture the step change."
]
},
{
"cell_type": "code",
"execution_count": 215,
"metadata": {
"id": "2DEc9aVF8WI6"
},
"outputs": [],
"source": [
"class LogLinkResNet2L(nn.Module):\n",
" # Define the parameters in __init__\n",
" def __init__(\n",
" self, \n",
" n_hidden, # size of the hidden layers \n",
" n_input=3, # number of inputs\n",
" n_output=1, # number of outputs\n",
" init_bias=0.0, # init mean value to speed up convergence \n",
" ): \n",
"\n",
" super(LogLinkResNet2L, self).__init__()\n",
"\n",
" self.hidden1 = torch.nn.Linear(n_input, n_hidden) # Hidden layer 1\n",
" self.batchn1 = torch.nn.BatchNorm1d(n_hidden) # Batchnorm layer \n",
"\n",
" self.hidden2 = torch.nn.Linear(n_hidden, n_hidden) # Hidden layer 2\n",
" self.batchn2 = torch.nn.BatchNorm1d(n_hidden) # Batchnorm layer \n",
"\n",
" self.linear = torch.nn.Linear(n_input, n_output) # Linear coefficients\n",
"\n",
" # Neural net (skip connections) coefficients\n",
" self.neural1 = torch.nn.Linear(n_hidden, n_output, bias=False) \n",
" self.neural2 = torch.nn.Linear(n_hidden, n_output, bias=False) \n",
"\n",
" nn.init.zeros_(self.linear.weight) # Initialise to zero\n",
" nn.init.zeros_(self.neural1.weight) # Initialise to zero\n",
" nn.init.zeros_(self.neural2.weight) # Initialise to zero\n",
" nn.init.constant_(self.linear.bias, init_bias) \n",
"\n",
"\n",
" # The forward function defines how you get y from X.\n",
" def forward(self, x):\n",
" h1 = F.relu(self.hidden1(x)) # Apply hidden layer \n",
" h1 = self.batchn1(h1) # Apply batchnorm \n",
"\n",
" h2 = torch.sigmoid(self.hidden2(h1)) # Apply hidden layer \n",
" h2 = self.batchn2(h2) # Apply batchnorm \n",
"\n",
" # Add GLM to NN(s)\n",
" lp = self.linear(x) + self.neural1(h1)+ self.neural2(h2)\n",
" return torch.exp(lp)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "b95VABNrf2DF"
},
"source": [
"## Applying models"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "09fWnXfCbJfE"
},
"source": [
"### A Note on Cross Validation\n",
"Onto fitting the deep learning models: the machine this is running on - and perhaps the readers machine - it takes a while to fit one model. So CV is going to take an even longer time. \n",
"\n",
"For a timesaving measure, we can try a higher learn rate and/or lesser epochs, for a significantly rougher guess at what the best parameters are. And since models are fairly similar (except for the 2 layer one) so we will pick n_hidden, l1 and weight decay based on the ResNet CV and apply the same values for the variations.\n",
"\n",
"Someone with a fast GPU may want to try this with lower ``max_lr`` and higher ``max_iter``."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "g8li21GRf6jh"
},
"source": [
"### LASSO\n",
"First, let us train the LASSO again:"
]
},
{
"cell_type": "code",
"execution_count": 230,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Ran in 641.7155 seconds\n"
]
},
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>mean_fit_time</th>\n",
" <th>std_fit_time</th>\n",
" <th>mean_score_time</th>\n",
" <th>std_score_time</th>\n",
" <th>param_l1_penalty</th>\n",
" <th>param_max_iter</th>\n",
" <th>param_verbose</th>\n",
" <th>params</th>\n",
" <th>split0_test_score</th>\n",
" <th>split1_test_score</th>\n",
" <th>split2_test_score</th>\n",
" <th>split3_test_score</th>\n",
" <th>split4_test_score</th>\n",
" <th>mean_test_score</th>\n",
" <th>std_test_score</th>\n",
" <th>rank_test_score</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>12</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 0.1, 'max_iter': 10, 'verbose': 0}</td>\n",
" <td>-39,617</td>\n",
" <td>-34,734</td>\n",
" <td>-35,648</td>\n",
" <td>-31,768</td>\n",
" <td>-28,989</td>\n",
" <td>-34,151</td>\n",
" <td>3,600</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>12</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 0.01, 'max_iter': 10, 'verbose'...</td>\n",
" <td>-39,617</td>\n",
" <td>-34,735</td>\n",
" <td>-35,648</td>\n",
" <td>-31,768</td>\n",
" <td>-28,989</td>\n",
" <td>-34,152</td>\n",
" <td>3,599</td>\n",
" <td>2</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>12</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 1.0, 'max_iter': 10, 'verbose': 0}</td>\n",
" <td>-39,617</td>\n",
" <td>-34,734</td>\n",
" <td>-35,651</td>\n",
" <td>-31,768</td>\n",
" <td>-28,990</td>\n",
" <td>-34,152</td>\n",
" <td>3,599</td>\n",
" <td>3</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>12</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 10.0, 'max_iter': 10, 'verbose'...</td>\n",
" <td>-39,616</td>\n",
" <td>-34,735</td>\n",
" <td>-35,653</td>\n",
" <td>-31,772</td>\n",
" <td>-28,995</td>\n",
" <td>-34,154</td>\n",
" <td>3,597</td>\n",
" <td>4</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>13</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 100.0, 'max_iter': 10, 'verbose...</td>\n",
" <td>-39,616</td>\n",
" <td>-34,737</td>\n",
" <td>-35,662</td>\n",
" <td>-31,782</td>\n",
" <td>-29,023</td>\n",
" <td>-34,164</td>\n",
" <td>3,589</td>\n",
" <td>5</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>12</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1,000</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 1000.0, 'max_iter': 10, 'verbos...</td>\n",
" <td>-39,612</td>\n",
" <td>-34,749</td>\n",
" <td>-35,687</td>\n",
" <td>-31,825</td>\n",
" <td>-29,047</td>\n",
" <td>-34,184</td>\n",
" <td>3,578</td>\n",
" <td>6</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8</th>\n",
" <td>13</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>50,000</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 50000.0, 'max_iter': 10, 'verbo...</td>\n",
" <td>-39,619</td>\n",
" <td>-34,784</td>\n",
" <td>-35,735</td>\n",
" <td>-31,880</td>\n",
" <td>-29,107</td>\n",
" <td>-34,225</td>\n",
" <td>3,560</td>\n",
" <td>7</td>\n",
" </tr>\n",
" <tr>\n",
" <th>7</th>\n",
" <td>12</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>20,000</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 20000.0, 'max_iter': 10, 'verbo...</td>\n",
" <td>-39,620</td>\n",
" <td>-34,786</td>\n",
" <td>-35,737</td>\n",
" <td>-31,882</td>\n",
" <td>-29,109</td>\n",
" <td>-34,227</td>\n",
" <td>3,560</td>\n",
" <td>8</td>\n",
" </tr>\n",
" <tr>\n",
" <th>6</th>\n",
" <td>12</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>7,674</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'l1_penalty': 7674.0, 'max_iter': 10, 'verbos...</td>\n",
" <td>-39,628</td>\n",
" <td>-34,797</td>\n",
" <td>-35,749</td>\n",
" <td>-31,895</td>\n",
" <td>-29,122</td>\n",
" <td>-34,238</td>\n",
" <td>3,558</td>\n",
" <td>9</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" mean_fit_time std_fit_time mean_score_time std_score_time \\\n",
"1 12 1 0 0 \n",
"0 12 0 0 0 \n",
"2 12 0 0 0 \n",
"3 12 0 0 0 \n",
"4 13 1 0 0 \n",
"5 12 0 0 0 \n",
"8 13 1 0 0 \n",
"7 12 0 0 0 \n",
"6 12 1 0 0 \n",
"\n",
" param_l1_penalty param_max_iter param_verbose \\\n",
"1 0 10 0 \n",
"0 0 10 0 \n",
"2 1 10 0 \n",
"3 10 10 0 \n",
"4 100 10 0 \n",
"5 1,000 10 0 \n",
"8 50,000 10 0 \n",
"7 20,000 10 0 \n",
"6 7,674 10 0 \n",
"\n",
" params split0_test_score \\\n",
"1 {'l1_penalty': 0.1, 'max_iter': 10, 'verbose': 0} -39,617 \n",
"0 {'l1_penalty': 0.01, 'max_iter': 10, 'verbose'... -39,617 \n",
"2 {'l1_penalty': 1.0, 'max_iter': 10, 'verbose': 0} -39,617 \n",
"3 {'l1_penalty': 10.0, 'max_iter': 10, 'verbose'... -39,616 \n",
"4 {'l1_penalty': 100.0, 'max_iter': 10, 'verbose... -39,616 \n",
"5 {'l1_penalty': 1000.0, 'max_iter': 10, 'verbos... -39,612 \n",
"8 {'l1_penalty': 50000.0, 'max_iter': 10, 'verbo... -39,619 \n",
"7 {'l1_penalty': 20000.0, 'max_iter': 10, 'verbo... -39,620 \n",
"6 {'l1_penalty': 7674.0, 'max_iter': 10, 'verbos... -39,628 \n",
"\n",
" split1_test_score split2_test_score split3_test_score split4_test_score \\\n",
"1 -34,734 -35,648 -31,768 -28,989 \n",
"0 -34,735 -35,648 -31,768 -28,989 \n",
"2 -34,734 -35,651 -31,768 -28,990 \n",
"3 -34,735 -35,653 -31,772 -28,995 \n",
"4 -34,737 -35,662 -31,782 -29,023 \n",
"5 -34,749 -35,687 -31,825 -29,047 \n",
"8 -34,784 -35,735 -31,880 -29,107 \n",
"7 -34,786 -35,737 -31,882 -29,109 \n",
"6 -34,797 -35,749 -31,895 -29,122 \n",
"\n",
" mean_test_score std_test_score rank_test_score \n",
"1 -34,151 3,600 1 \n",
"0 -34,152 3,599 2 \n",
"2 -34,152 3,599 3 \n",
"3 -34,154 3,597 4 \n",
"4 -34,164 3,589 5 \n",
"5 -34,184 3,578 6 \n",
"8 -34,225 3,560 7 \n",
"7 -34,227 3,560 8 \n",
"6 -34,238 3,558 9 "
]
},
"execution_count": 230,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tic = time.perf_counter() # Timing\n",
"lasso_CV = Pipeline(\n",
" steps=[\n",
" (\"keep\", ColumnKeeper(list_of_features)),\n",
" (\"transform\", RampTransformer(num_periods=num_periods)), \n",
" (\"model\", GridSearchCV(\n",
" TabularNetRegressor(LogLinkGLM),\n",
" {\n",
" 'verbose': [0],\n",
" 'max_iter': [10], # Increase this for better results\n",
" 'l1_penalty': [0.01, 0.1, 1.0, 10.0, 100.0, 1000.0, 7674.0, 20000.0, 50000.0],\n",
" },\n",
" cv=5,\n",
" error_score=-np.Inf # Consider errors/non-convergence to be worst case\n",
" )\n",
" )\n",
" ]\n",
")\n",
"lasso_CV.fit(X_train, y_train)\n",
"toc = time.perf_counter()\n",
"\n",
"print(f\"Ran in {toc - tic:0.4f} seconds\")\n",
"\n",
"pd.DataFrame(lasso_CV[\"model\"].cv_results_).sort_values(\"rank_test_score\") # Results"
]
},
{
"cell_type": "code",
"execution_count": 231,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "P79wt--s9ym_",
"outputId": "b293ae9c-2313-41c8-b0a7-e3c9dda5b333"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train RMSE: 34417.40234375 Train Loss: -21047.498046875\n",
"Train RMSE: 34365.20703125 Train Loss: -21664.224609375\n",
"Train RMSE: 34320.4296875 Train Loss: -22303.412109375\n",
"Train RMSE: 34305.09375 Train Loss: -22412.63671875\n",
"Train RMSE: 34296.796875 Train Loss: -22516.00390625\n",
"Train RMSE: 34293.421875 Train Loss: -22550.36328125\n",
"Train RMSE: 34290.6328125 Train Loss: -22571.49609375\n",
"Train RMSE: 34289.01953125 Train Loss: -22583.134765625\n",
"Train RMSE: 34288.3828125 Train Loss: -22589.09375\n",
"Train RMSE: 34288.16796875 Train Loss: -22591.3515625\n",
"Ran in 101.7433 seconds\n"
]
}
],
"source": [
"tic = time.perf_counter()\n",
"\n",
"lasso = Pipeline(\n",
" steps=[\n",
" (\"keep\", ColumnKeeper(list_of_features)),\n",
" (\"transform\", RampTransformer(num_periods=num_periods)), \n",
" (\"model\", TabularNetRegressor(\n",
" module=LogLinkGLM, \n",
" l1_penalty=lasso_CV[\"model\"].best_params_[\"l1_penalty\"]))\n",
" ]\n",
")\n",
"\n",
"lasso.fit(X_train, y_train)\n",
"\n",
"toc = time.perf_counter()\n",
"print(f\"Ran in {toc - tic:0.4f} seconds\")"
]
},
{
"cell_type": "code",
"execution_count": 232,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"array([4394.9995 , 4418.8584 , 4442.97 , ..., 48.56727 ,\n",
" 39.500504, 32.04342 ], dtype=float32)"
]
},
"execution_count": 232,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"lasso.predict(X_train)"
]
},
{
"cell_type": "code",
"execution_count": 233,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "uj3SR_jHHL0C",
"outputId": "02ee81ef-475f-4084-e0ed-7c0183dbf5a7"
},
"outputs": [
{
"data": {
"text/plain": [
"-30477.82913629737"
]
},
"execution_count": 233,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# Should give good results on test data:\n",
"lasso.score(X_test, y_test)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Next, use CV to search for a good ResNet model:"
]
},
{
"cell_type": "code",
"execution_count": 234,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "o3gJszD65I6_",
"outputId": "7ffba9cb-41c9-43c4-cf00-674a9d944e1f"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Ran in 247.2704 seconds\n"
]
},
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>mean_fit_time</th>\n",
" <th>std_fit_time</th>\n",
" <th>mean_score_time</th>\n",
" <th>std_score_time</th>\n",
" <th>param_weight_decay</th>\n",
" <th>param_verbose</th>\n",
" <th>param_n_hidden</th>\n",
" <th>param_max_iter</th>\n",
" <th>param_l1_penalty</th>\n",
" <th>params</th>\n",
" <th>split0_test_score</th>\n",
" <th>split1_test_score</th>\n",
" <th>split2_test_score</th>\n",
" <th>split3_test_score</th>\n",
" <th>split4_test_score</th>\n",
" <th>mean_test_score</th>\n",
" <th>std_test_score</th>\n",
" <th>rank_test_score</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>6</th>\n",
" <td>5</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1000</td>\n",
" <td>10</td>\n",
" <td>1</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,617</td>\n",
" <td>-34,765</td>\n",
" <td>-35,712</td>\n",
" <td>-31,861</td>\n",
" <td>-29,096</td>\n",
" <td>-34,210</td>\n",
" <td>3,563</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>19</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>1,000</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,613</td>\n",
" <td>-34,775</td>\n",
" <td>-35,726</td>\n",
" <td>-31,865</td>\n",
" <td>-29,093</td>\n",
" <td>-34,214</td>\n",
" <td>3,564</td>\n",
" <td>2</td>\n",
" </tr>\n",
" <tr>\n",
" <th>13</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>250</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'weight_decay': 0.01, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,617</td>\n",
" <td>-34,774</td>\n",
" <td>-35,724</td>\n",
" <td>-31,861</td>\n",
" <td>-29,096</td>\n",
" <td>-34,215</td>\n",
" <td>3,564</td>\n",
" <td>3</td>\n",
" </tr>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>250</td>\n",
" <td>10</td>\n",
" <td>1,000</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,616</td>\n",
" <td>-34,776</td>\n",
" <td>-35,728</td>\n",
" <td>-31,871</td>\n",
" <td>-29,098</td>\n",
" <td>-34,218</td>\n",
" <td>3,563</td>\n",
" <td>4</td>\n",
" </tr>\n",
" <tr>\n",
" <th>5</th>\n",
" <td>3</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>500</td>\n",
" <td>10</td>\n",
" <td>10</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,621</td>\n",
" <td>-34,779</td>\n",
" <td>-35,728</td>\n",
" <td>-31,876</td>\n",
" <td>-29,099</td>\n",
" <td>-34,220</td>\n",
" <td>3,563</td>\n",
" <td>5</td>\n",
" </tr>\n",
" <tr>\n",
" <th>7</th>\n",
" <td>3</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>500</td>\n",
" <td>10</td>\n",
" <td>20,000</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,620</td>\n",
" <td>-34,783</td>\n",
" <td>-35,733</td>\n",
" <td>-31,880</td>\n",
" <td>-29,108</td>\n",
" <td>-34,225</td>\n",
" <td>3,560</td>\n",
" <td>6</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>50,000</td>\n",
" <td>{'weight_decay': 0.01, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,625</td>\n",
" <td>-34,793</td>\n",
" <td>-35,742</td>\n",
" <td>-31,888</td>\n",
" <td>-29,117</td>\n",
" <td>-34,233</td>\n",
" <td>3,559</td>\n",
" <td>7</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>250</td>\n",
" <td>10</td>\n",
" <td>7,674</td>\n",
" <td>{'weight_decay': 0.01, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,627</td>\n",
" <td>-34,794</td>\n",
" <td>-35,745</td>\n",
" <td>-31,891</td>\n",
" <td>-29,122</td>\n",
" <td>-34,236</td>\n",
" <td>3,558</td>\n",
" <td>8</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>250</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'weight_decay': 1.0, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,637</td>\n",
" <td>-34,787</td>\n",
" <td>-35,751</td>\n",
" <td>-31,891</td>\n",
" <td>-29,125</td>\n",
" <td>-34,238</td>\n",
" <td>3,561</td>\n",
" <td>9</td>\n",
" </tr>\n",
" <tr>\n",
" <th>18</th>\n",
" <td>6</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>1000</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'weight_decay': 1.0, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,643</td>\n",
" <td>-34,793</td>\n",
" <td>-35,744</td>\n",
" <td>-31,884</td>\n",
" <td>-29,135</td>\n",
" <td>-34,240</td>\n",
" <td>3,560</td>\n",
" <td>10</td>\n",
" </tr>\n",
" <tr>\n",
" <th>15</th>\n",
" <td>3</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>500</td>\n",
" <td>10</td>\n",
" <td>100</td>\n",
" <td>{'weight_decay': 1.0, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,641</td>\n",
" <td>-34,794</td>\n",
" <td>-35,751</td>\n",
" <td>-31,894</td>\n",
" <td>-29,119</td>\n",
" <td>-34,240</td>\n",
" <td>3,563</td>\n",
" <td>11</td>\n",
" </tr>\n",
" <tr>\n",
" <th>8</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>50,000</td>\n",
" <td>{'weight_decay': 1.0, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,638</td>\n",
" <td>-34,801</td>\n",
" <td>-35,756</td>\n",
" <td>-31,900</td>\n",
" <td>-29,125</td>\n",
" <td>-34,244</td>\n",
" <td>3,561</td>\n",
" <td>12</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>5</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>1000</td>\n",
" <td>10</td>\n",
" <td>7,674</td>\n",
" <td>{'weight_decay': 0.1, 'verbose': 0, 'n_hidden'...</td>\n",
" <td>-39,635</td>\n",
" <td>-34,802</td>\n",
" <td>-35,755</td>\n",
" <td>-31,901</td>\n",
" <td>-29,137</td>\n",
" <td>-34,246</td>\n",
" <td>3,556</td>\n",
" <td>13</td>\n",
" </tr>\n",
" <tr>\n",
" <th>10</th>\n",
" <td>3</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>500</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,739</td>\n",
" <td>-34,897</td>\n",
" <td>-35,860</td>\n",
" <td>-32,011</td>\n",
" <td>-29,236</td>\n",
" <td>-34,348</td>\n",
" <td>3,557</td>\n",
" <td>14</td>\n",
" </tr>\n",
" <tr>\n",
" <th>9</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>250</td>\n",
" <td>10</td>\n",
" <td>1,000</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,741</td>\n",
" <td>-34,902</td>\n",
" <td>-35,865</td>\n",
" <td>-32,016</td>\n",
" <td>-29,239</td>\n",
" <td>-34,353</td>\n",
" <td>3,556</td>\n",
" <td>15</td>\n",
" </tr>\n",
" <tr>\n",
" <th>14</th>\n",
" <td>5</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>1000</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,743</td>\n",
" <td>-34,903</td>\n",
" <td>-35,865</td>\n",
" <td>-32,015</td>\n",
" <td>-29,240</td>\n",
" <td>-34,353</td>\n",
" <td>3,557</td>\n",
" <td>16</td>\n",
" </tr>\n",
" <tr>\n",
" <th>12</th>\n",
" <td>5</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>1000</td>\n",
" <td>10</td>\n",
" <td>20,000</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,743</td>\n",
" <td>-34,904</td>\n",
" <td>-35,867</td>\n",
" <td>-32,018</td>\n",
" <td>-29,241</td>\n",
" <td>-34,354</td>\n",
" <td>3,556</td>\n",
" <td>17</td>\n",
" </tr>\n",
" <tr>\n",
" <th>11</th>\n",
" <td>3</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>500</td>\n",
" <td>10</td>\n",
" <td>20,000</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,743</td>\n",
" <td>-34,904</td>\n",
" <td>-35,867</td>\n",
" <td>-32,018</td>\n",
" <td>-29,242</td>\n",
" <td>-34,354</td>\n",
" <td>3,556</td>\n",
" <td>18</td>\n",
" </tr>\n",
" <tr>\n",
" <th>17</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>10</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>20,000</td>\n",
" <td>{'weight_decay': 10.0, 'verbose': 0, 'n_hidden...</td>\n",
" <td>-39,743</td>\n",
" <td>-34,904</td>\n",
" <td>-35,867</td>\n",
" <td>-32,018</td>\n",
" <td>-29,242</td>\n",
" <td>-34,355</td>\n",
" <td>3,556</td>\n",
" <td>19</td>\n",
" </tr>\n",
" <tr>\n",
" <th>16</th>\n",
" <td>1</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>0</td>\n",
" <td>100</td>\n",
" <td>10</td>\n",
" <td>50,000</td>\n",
" <td>{'weight_decay': 100.0, 'verbose': 0, 'n_hidde...</td>\n",
" <td>-39,754</td>\n",
" <td>-34,916</td>\n",
" <td>-35,880</td>\n",
" <td>-32,031</td>\n",
" <td>-29,256</td>\n",
" <td>-34,367</td>\n",
" <td>3,555</td>\n",
" <td>20</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" mean_fit_time std_fit_time mean_score_time std_score_time \\\n",
"6 5 0 0 0 \n",
"19 1 0 0 0 \n",
"13 1 0 0 0 \n",
"0 1 0 0 0 \n",
"5 3 0 0 0 \n",
"7 3 0 0 0 \n",
"3 1 0 0 0 \n",
"2 1 0 0 0 \n",
"1 1 0 0 0 \n",
"18 6 0 0 0 \n",
"15 3 0 0 0 \n",
"8 1 0 0 0 \n",
"4 5 0 0 0 \n",
"10 3 0 0 0 \n",
"9 1 0 0 0 \n",
"14 5 0 0 0 \n",
"12 5 0 0 0 \n",
"11 3 0 0 0 \n",
"17 1 0 0 0 \n",
"16 1 0 0 0 \n",
"\n",
" param_weight_decay param_verbose param_n_hidden param_max_iter \\\n",
"6 0 0 1000 10 \n",
"19 0 0 100 10 \n",
"13 0 0 250 10 \n",
"0 0 0 250 10 \n",
"5 0 0 500 10 \n",
"7 0 0 500 10 \n",
"3 0 0 100 10 \n",
"2 0 0 250 10 \n",
"1 1 0 250 10 \n",
"18 1 0 1000 10 \n",
"15 1 0 500 10 \n",
"8 1 0 100 10 \n",
"4 0 0 1000 10 \n",
"10 10 0 500 10 \n",
"9 10 0 250 10 \n",
"14 10 0 1000 10 \n",
"12 10 0 1000 10 \n",
"11 10 0 500 10 \n",
"17 10 0 100 10 \n",
"16 100 0 100 10 \n",
"\n",
" param_l1_penalty params \\\n",
"6 1 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"19 1,000 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"13 0 {'weight_decay': 0.01, 'verbose': 0, 'n_hidden... \n",
"0 1,000 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"5 10 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"7 20,000 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"3 50,000 {'weight_decay': 0.01, 'verbose': 0, 'n_hidden... \n",
"2 7,674 {'weight_decay': 0.01, 'verbose': 0, 'n_hidden... \n",
"1 0 {'weight_decay': 1.0, 'verbose': 0, 'n_hidden'... \n",
"18 0 {'weight_decay': 1.0, 'verbose': 0, 'n_hidden'... \n",
"15 100 {'weight_decay': 1.0, 'verbose': 0, 'n_hidden'... \n",
"8 50,000 {'weight_decay': 1.0, 'verbose': 0, 'n_hidden'... \n",
"4 7,674 {'weight_decay': 0.1, 'verbose': 0, 'n_hidden'... \n",
"10 0 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"9 1,000 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"14 0 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"12 20,000 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"11 20,000 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"17 20,000 {'weight_decay': 10.0, 'verbose': 0, 'n_hidden... \n",
"16 50,000 {'weight_decay': 100.0, 'verbose': 0, 'n_hidde... \n",
"\n",
" split0_test_score split1_test_score split2_test_score \\\n",
"6 -39,617 -34,765 -35,712 \n",
"19 -39,613 -34,775 -35,726 \n",
"13 -39,617 -34,774 -35,724 \n",
"0 -39,616 -34,776 -35,728 \n",
"5 -39,621 -34,779 -35,728 \n",
"7 -39,620 -34,783 -35,733 \n",
"3 -39,625 -34,793 -35,742 \n",
"2 -39,627 -34,794 -35,745 \n",
"1 -39,637 -34,787 -35,751 \n",
"18 -39,643 -34,793 -35,744 \n",
"15 -39,641 -34,794 -35,751 \n",
"8 -39,638 -34,801 -35,756 \n",
"4 -39,635 -34,802 -35,755 \n",
"10 -39,739 -34,897 -35,860 \n",
"9 -39,741 -34,902 -35,865 \n",
"14 -39,743 -34,903 -35,865 \n",
"12 -39,743 -34,904 -35,867 \n",
"11 -39,743 -34,904 -35,867 \n",
"17 -39,743 -34,904 -35,867 \n",
"16 -39,754 -34,916 -35,880 \n",
"\n",
" split3_test_score split4_test_score mean_test_score std_test_score \\\n",
"6 -31,861 -29,096 -34,210 3,563 \n",
"19 -31,865 -29,093 -34,214 3,564 \n",
"13 -31,861 -29,096 -34,215 3,564 \n",
"0 -31,871 -29,098 -34,218 3,563 \n",
"5 -31,876 -29,099 -34,220 3,563 \n",
"7 -31,880 -29,108 -34,225 3,560 \n",
"3 -31,888 -29,117 -34,233 3,559 \n",
"2 -31,891 -29,122 -34,236 3,558 \n",
"1 -31,891 -29,125 -34,238 3,561 \n",
"18 -31,884 -29,135 -34,240 3,560 \n",
"15 -31,894 -29,119 -34,240 3,563 \n",
"8 -31,900 -29,125 -34,244 3,561 \n",
"4 -31,901 -29,137 -34,246 3,556 \n",
"10 -32,011 -29,236 -34,348 3,557 \n",
"9 -32,016 -29,239 -34,353 3,556 \n",
"14 -32,015 -29,240 -34,353 3,557 \n",
"12 -32,018 -29,241 -34,354 3,556 \n",
"11 -32,018 -29,242 -34,354 3,556 \n",
"17 -32,018 -29,242 -34,355 3,556 \n",
"16 -32,031 -29,256 -34,367 3,555 \n",
"\n",
" rank_test_score \n",
"6 1 \n",
"19 2 \n",
"13 3 \n",
"0 4 \n",
"5 5 \n",
"7 6 \n",
"3 7 \n",
"2 8 \n",
"1 9 \n",
"18 10 \n",
"15 11 \n",
"8 12 \n",
"4 13 \n",
"10 14 \n",
"9 15 \n",
"14 16 \n",
"12 17 \n",
"11 18 \n",
"17 19 \n",
"16 20 "
]
},
"execution_count": 234,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"tic = time.perf_counter() # Timing\n",
"res_net_CV = Pipeline(\n",
" steps=[\n",
" (\"keep\", ColumnKeeper(list_of_features)), \n",
" ('zero_to_one', MinMaxScaler()), # Important! Standardize deep learning inputs.\n",
" (\"model\", RandomizedSearchCV(\n",
" TabularNetRegressor(LogLinkResNet),\n",
" {\n",
" 'verbose': [0],\n",
" 'max_iter': [10], # Increase this for better results\n",
" 'n_hidden': [100, 250, 500, 1000],\n",
" 'l1_penalty': [0.01, 0.1, 1.0, 10.0, 100.0, 1000.0, 7674.0, 20000.0, 50000.0],\n",
" 'weight_decay': [0.01, 0.1, 1.0, 10.0, 100.0],\n",
" },\n",
" n_iter=20,\n",
" cv=5,\n",
" error_score=-np.Inf, # Consider errors/non-convergence to be worst case\n",
" random_state=0)\n",
" )\n",
" ]\n",
")\n",
"res_net_CV.fit(X_train, y_train)\n",
"toc = time.perf_counter()\n",
"\n",
"print(f\"Ran in {toc - tic:0.4f} seconds\")\n",
"\n",
"pd.DataFrame(res_net_CV[\"model\"].cv_results_).sort_values(\"rank_test_score\") # Results"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "qYjVt7nycHVK"
},
"source": [
"The best parameters were, or are (if you ran the CV) as follows:"
]
},
{
"cell_type": "code",
"execution_count": 235,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "niFQ72H-jwf5",
"outputId": "7370e3d0-b7e8-4a21-dbe0-018bdb41e94a"
},
"outputs": [
{
"data": {
"text/plain": [
"{'weight_decay': 0.1,\n",
" 'verbose': 0,\n",
" 'n_hidden': 1000,\n",
" 'max_iter': 10,\n",
" 'l1_penalty': 1.0}"
]
},
"execution_count": 235,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"try: \n",
" best = res_net_CV[\"model\"].best_params_\n",
" \n",
"except: \n",
" \n",
" # Previous CV run was as follows:\n",
" best = {\n",
" 'weight_decay': 0.1,\n",
" 'verbose': 0,\n",
" 'n_hidden': 1000,\n",
" 'l1_penalty': 1.0\n",
" }\n",
" \n",
"best "
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "GzN385n7aCA7"
},
"source": [
"Next, the feed-forward net, ResNet and DenseNet. For brevity and simplicity, we will use the same parameters as the best parameters from the ResNet CV, and put the logic into a loop across all three models."
]
},
{
"cell_type": "code",
"execution_count": 236,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "CS7vty_Ikrze",
"outputId": "9d7e24ca-d21b-4c87-d248-87ba0ded881f"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Train RMSE: 34416.57421875 Train Loss: -21047.498046875\n",
"Train RMSE: 34399.828125 Train Loss: -22202.482421875\n",
"Train RMSE: 34374.40625 Train Loss: -22456.8984375\n",
"Train RMSE: 34387.57421875 Train Loss: -22562.65234375\n",
"Train RMSE: 34343.94140625 Train Loss: -22751.435546875\n",
"Train RMSE: 34317.2265625 Train Loss: -22812.666015625\n",
"Train RMSE: 34296.7109375 Train Loss: -22830.560546875\n",
"Train RMSE: 34282.5390625 Train Loss: -22838.400390625\n",
"Train RMSE: 34271.9453125 Train Loss: -22842.984375\n",
"Train RMSE: 34262.796875 Train Loss: -22844.923828125\n",
"Ran in 79.1514 seconds\n",
"Train RMSE: 34416.62109375 Train Loss: -21047.498046875\n",
"Train RMSE: 34399.54296875 Train Loss: -22206.767578125\n",
"Train RMSE: 34375.9609375 Train Loss: -22420.6796875\n",
"Train RMSE: 34381.54296875 Train Loss: -22469.541015625\n",
"Train RMSE: 34336.4453125 Train Loss: -22778.904296875\n",
"Train RMSE: 34313.30078125 Train Loss: -22812.52734375\n",
"Train RMSE: 34295.69140625 Train Loss: -22827.986328125\n",
"Train RMSE: 34283.2578125 Train Loss: -22838.48046875\n",
"Train RMSE: 34271.265625 Train Loss: -22844.029296875\n",
"Train RMSE: 34262.32421875 Train Loss: -22846.52734375\n",
"Ran in 83.7604 seconds\n",
"Train RMSE: 34416.51171875 Train Loss: -21047.498046875\n",
"Train RMSE: 34399.13671875 Train Loss: -22214.091796875\n",
"Train RMSE: 34377.6953125 Train Loss: -22419.826171875\n",
"Train RMSE: 34387.24609375 Train Loss: -22705.12890625\n",
"Train RMSE: 34339.16015625 Train Loss: -22772.21875\n",
"Train RMSE: 34313.46875 Train Loss: -22812.8203125\n",
"Train RMSE: 34297.09765625 Train Loss: -22825.677734375\n",
"Train RMSE: 34283.69921875 Train Loss: -22833.203125\n",
"Train RMSE: 34272.06640625 Train Loss: -22837.109375\n",
"Train RMSE: 34262.72265625 Train Loss: -22838.966796875\n",
"Ran in 226.1781 seconds\n",
"Train RMSE: 34415.9453125 Train Loss: -21047.498046875\n",
"Train RMSE: 34397.16796875 Train Loss: -22335.376953125\n",
"Train RMSE: 34416.01953125 Train Loss: -21281.458984375\n",
"Train RMSE: 34526.28515625 Train Loss: -17113.123046875\n",
"Train RMSE: 34351.66015625 Train Loss: -21575.189453125\n",
"Train RMSE: 34318.05078125 Train Loss: -22078.115234375\n",
"Train RMSE: 34300.99609375 Train Loss: -22597.1640625\n",
"Train RMSE: 34290.640625 Train Loss: -22770.873046875\n",
"Train RMSE: 34279.94140625 Train Loss: -22787.98046875\n",
"Train RMSE: 34271.43359375 Train Loss: -22791.75390625\n",
"Ran in 323.8305 seconds\n"
]
}
],
"source": [
"models = []\n",
"\n",
"for Model in [LogLinkForwardNet, LogLinkResNet, LogLinkDenseNet, LogLinkResNet2L]:\n",
" tic = time.perf_counter() # Timing\n",
"\n",
" net = Pipeline(\n",
" steps=[\n",
" (\"keep\", ColumnKeeper(list_of_features)),\n",
" ('zero_to_one', MinMaxScaler()), # Important! Standardize deep learning inputs.\n",
" (\"model\", \n",
" TabularNetRegressor(\n",
" module=Model,\n",
" n_hidden=best[\"n_hidden\"],\n",
" l1_penalty=best[\"l1_penalty\"],\n",
" weight_decay=best[\"weight_decay\"]\n",
" )\n",
" )\n",
" ]\n",
" )\n",
"\n",
" net.fit(X_train, y_train)\n",
"\n",
" toc = time.perf_counter()\n",
" print(f\"Ran in {toc - tic:0.4f} seconds\")\n",
"\n",
" models += [net]\n",
"\n",
"fwd_net, res_net, dense_net, res_net2 = models"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "p6aZtMX3BPNX"
},
"source": [
"## Results"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### RMSE"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "NOQlu4eFcPeT"
},
"source": [
"Calculate test RMSE for each model:"
]
},
{
"cell_type": "code",
"execution_count": 237,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "yVy146NDZ3qD",
"outputId": "26356090-8e71-45a6-e5ec-7a1db6aa15b6"
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Lasso\n",
"Feedforward\n",
"ResNet\n",
"DenseNet\n",
"ResNet (2 Layers)\n"
]
}
],
"source": [
"# Empty list to store results.\n",
"scikit_results = []\n",
"y_predicted_full_results = {}\n",
"y_predicted_test_results = {\"Actuals\": y_test}\n",
"\n",
"models = [lasso, fwd_net, res_net, dense_net, res_net2]\n",
"model_names = [\"Lasso\", \"Feedforward\", \"ResNet\", \"DenseNet\", \"ResNet (2 Layers)\"]\n",
"\n",
"# Using zip in a for loop means pipe and name will run through \n",
"# the tuples of the two lists in that order.\n",
"for pipe, name in zip(models, model_names):\n",
" print(name)\n",
" y_predicted_train = pipe.predict(X_train)\n",
" train_rmse = mean_squared_error(y_train, y_predicted_train, squared=False)\n",
"\n",
" y_predicted_test = pipe.predict(X_test)\n",
" test_rmse = mean_squared_error(y_test, y_predicted_test, squared=False)\n",
" \n",
" y_predicted_full_results[name] = pipe.predict(X)\n",
" y_predicted_test_results[name] = y_predicted_test\n",
" \n",
" scikit_results += [{\n",
" \"Name\": name, \n",
" \"Train RMSE\": train_rmse,\n",
" \"Test RMSE\": test_rmse\n",
" }]"
]
},
{
"cell_type": "code",
"execution_count": 238,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/",
"height": 237
},
"id": "IWSXnoU6aYJ-",
"outputId": "7131ebd8-0431-47cd-846b-36bf6527a1f4"
},
"outputs": [
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>Name</th>\n",
" <th>Train RMSE</th>\n",
" <th>Test RMSE</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>DenseNet</td>\n",
" <td>34,256</td>\n",
" <td>30,455</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>Feedforward</td>\n",
" <td>34,256</td>\n",
" <td>30,456</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>ResNet</td>\n",
" <td>34,256</td>\n",
" <td>30,457</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>ResNet (2 Layers)</td>\n",
" <td>34,265</td>\n",
" <td>30,457</td>\n",
" </tr>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>Lasso</td>\n",
" <td>34,288</td>\n",
" <td>30,478</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" Name Train RMSE Test RMSE\n",
"3 DenseNet 34,256 30,455\n",
"1 Feedforward 34,256 30,456\n",
"2 ResNet 34,256 30,457\n",
"4 ResNet (2 Layers) 34,265 30,457\n",
"0 Lasso 34,288 30,478"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"df_results = pd.DataFrame(scikit_results).sort_values(\"Test RMSE\")\n",
"pd.set_option('display.float_format', '{0:,.0f}'.format)\n",
"display(df_results)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"The predictions are all fairly similar in both Train and Test RMSE."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Plots\n",
"Here are some quick plots using just the ``plot`` function in ``pandas``. For nicer plots, the previous article has the ``QTrack`` plot functions which can also be applied."
]
},
{
"cell_type": "code",
"execution_count": 239,
"metadata": {},
"outputs": [],
"source": [
"model_forecasts = (\n",
" pd.concat(\n",
" [dat, pd.DataFrame(y_predicted_full_results)],\n",
" axis='columns'\n",
" )\n",
" .groupby([\"occurrence_period\", \"development_period\"])\n",
" .agg('sum')\n",
" .reset_index()\n",
")"
]
},
{
"cell_type": "code",
"execution_count": 240,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x16c1fba60>"
]
},
"execution_count": 240,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Lasso - fits quite well (except with a hook on the tail)\n",
"model_forecasts.pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"Lasso\").plot(logy=True)\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "code",
"execution_count": 241,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x16c1fba00>"
]
},
"execution_count": 241,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# Feedforward\n",
"model_forecasts.pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"Feedforward\").plot(logy=True)\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "code",
"execution_count": 242,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x16c312cd0>"
]
},
"execution_count": 242,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# ResNet\n",
"model_forecasts.pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"ResNet\").plot(logy=True)\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "code",
"execution_count": 243,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x16ca4d370>"
]
},
"execution_count": 243,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# DenseNet\n",
"model_forecasts.pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"DenseNet\").plot(logy=True)\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "code",
"execution_count": 244,
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"<matplotlib.legend.Legend at 0x16caddaf0>"
]
},
"execution_count": 244,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# ResNet 2 layers\n",
"model_forecasts.pivot(index=\"development_period\", columns=\"occurrence_period\", values=\"ResNet (2 Layers)\").plot(logy=True)\n",
"plt.legend(loc=\"lower center\", bbox_to_anchor=(0.5, -0.8), ncol=5)"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "sX_V24tWMjgx"
},
"source": [
"All of the models seem fairly similar in terms of shape. The LASSO seems to have the smoothest curve whilst the 2 layer model may be overfitting somewhat."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "9lg8NNcpB8I2"
},
"source": [
"## Conclusion"
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "sX_V24tWMjgx"
},
"source": [
"In this article we delved into the feedforward, ResNet and DenseNet neural networks structures, and looked at how they can be applied to individual claims reserving regression problems with GLM-style link functions."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "sX_V24tWMjgx"
},
"source": [
"There are more advanced architectures like [\"Deep Neural Decision Trees\"](https://arxiv.org/pdf/1806.06988.pdf) or [TabNet](https://arxiv.org/pdf/1908.07442.pdf) that may or may not work better for this. There are also more sophisticated techniques in literature that we have not covered here, including [embeddings](https://www.fast.ai/2018/04/29/categorical-embeddings/), [dropout](https://machinelearningmastery.com/dropout-for-regularizing-deep-neural-networks/), and models that cater specifically to reserving, such as [DeepTriangle](https://arxiv.org/abs/1804.09253) and [ResMDN](https://institute-and-faculty-of-actuaries.github.io/mlr-blog/post/l-nn-al-mudafer/) models."
]
},
{
"cell_type": "markdown",
"metadata": {
"id": "sX_V24tWMjgx"
},
"source": [
"An important call out is that generally, with smaller datasets, simpler models usually works better. The opportunities for machine learning models to shine with this dataset design are still quite limited. Here we have only a limited number of features. More sophisticated machine learning models such as deep learning often only get their time to shine when there is richer data. For claims reserving this could be by enhancing the dataset with additional claims-level, granular data."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"However we hope this provides a good overview. Interested readers could further explore some of the ideas raised:\n",
" * How would a \"snapshot-based\" tabular structure (see the \"Limitations of this data structure\") perform?\n",
" * How would RNN based models perform on this dataset?\n",
" * Would a neural network with multiple outputs to predict both payments and settlement outperform?\n",
" * How do different models perform with different train/test/cross-validation techniques?\n",
" * How would probabilistic models perform?\n",
" * Whilst this model predicts outstanding case estimates / IBNER, how could an IBNR model be established?"
]
}
],
"metadata": {
"accelerator": "GPU",
"colab": {
"collapsed_sections": [],
"name": "Tabular Neural Networks for Triangles.ipynb",
"provenance": [],
"toc_visible": true
},
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.9.7"
}
},
"nbformat": 4,
"nbformat_minor": 1
}
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment