cereniyim

@dataclass
class ModelMetaData:
    estimator: IsolationForest
    model_path: str


class AnomalyDetector:
    def __init__(self, estimator=None):
        self._estimator = estimator or IsolationForest(
            random_state=42,

cereniyim

#!/usr/bin/env bash
if test -z "$1"
then
      echo "Usage ./build-task-images.sh VERSION"
      echo "No version was passed! Please pass a version to the script e.g. 0.1"
      exit 1
fi

VERSION=$1
docker build -t code-challenge/base-docker base_docker

cereniyim

import pandas as pd
import numpy as np
import click
from pathlib import Path
from sklearn.ensemble import RandomForestRegressor
import pickle

def convert_features_to_array(features):
    num_rows = len(features)
    num_cols = len(features.columns)

cereniyim

version: '2'
services:

  luigid:
    image: code-challenge/orchestrator:latest
    build:
      context: ./orchestrator
    command: luigid
    ports:
      - "8082:8082"

cereniyim

import luigi
import os
from pathlib import Path

from util import DockerTask

class TrainModel(DockerTask):
    """Task to train random forest regressor with training datasets"""

    in_path = '/usr/share/data/interim/'

cereniyim

class MakeDatasets(DockerTask):
    """Task to separate training and test sets"""

    in_csv = luigi.Parameter(default='/usr/share/data/raw/wine_dataset.csv')
    out_dir = luigi.Parameter(default='/usr/share/data/interim/')
    flag = luigi.Parameter('.SUCCESS_MakeDatasets')

    @property
    def image(self):
        return f'code-challenge/make-dataset:{VERSION}'

cereniyim

def PlotFeatureImportances(model, feature_names):
    feature_importances = (pd
                           .DataFrame(
                               {'feature': feature_names,
                                'importance': model
                                .feature_importances_}))

    feature_importances = (feature_importances
                           .sort_values(by="importance",
                                        ascending=False))

cereniyim

def PlotPredictedVSActual(predictions, actuals):
    figsize(20, 10)
    plt.rcParams['font.size'] = 14

    # histogram of predictions
    ax = plt.subplot(121)
    ax.hist(predictions,
            bins=10, 
            color = "#971539", 
            edgecolor = 'white')

cereniyim

# visualization and manipulation libraries
import matplotlib.pyplot as plt
import seaborn as sns
from IPython.core.pylabtools import figsize
import pandas as pd
import numpy as np
pd.set_option('display.max_columns', 60)

%matplotlib inline
# to display visuals in the notebook

cereniyim

def fit_evaluate_model(model, X_train, y_train, 
                       X_valid, y_valid):
    # function to train a given model
    # return mean squared error of the
    # actuals and predictions
    model.fit(X_train, y_train)
    y_predicted = model.predict(X_valid)
    return mean_squared_error(y_valid, y_predicted)