eddjberry

#=================================================
# geom_line() + geom_ribbon()
#=================================================
# plots by group
plot_by_group <- function(df, x, colour) {
  # create the summary data using # group_prop()
  df_summary <- df %>%
    dplyr::filter(!is.na({{ colour }})) %>%
    group_prop({{ x }}, {{ colour }})
  

eddjberry

group_prop <- function(df, ...) {
  # enquo the dots
  vars <- enquos(...)
  
  # count then calculate
  # proportions
  df_count <- df %>%
    count(!!!vars)
  
  if (length(vars) > 1) {

eddjberry

def partial_dependency_data(df, model, col, values, sample_fraction = 0.1):
  
  # empty list for predictions
  avg_predictions = list()

  # take a sample of the data to use
  df_sample = df.sample(fraction = sample_fraction)
  
  # loop through the values
  for val in values:

eddjberry

#!/usr/bin/env Rscript

# to run from command line:
## chmod +x knit_dir.R
## ./knit_dir.R <dir-name>

# from https://stackoverflow.com/a/49950761
# to avoid conflicts between packages
# breaking things
clean_search <- function() {

eddjberry

import pandas as pd
from faker import Faker

# set the seed
Faker.seed(10)

# set the locale to GB
fake = Faker("en_GB")

# how many customers to fake

eddjberry

import timeit
import numpy as np
from faker import Faker

# create the faker object
fake = Faker()

# np.random_choice function
def np_choice(N=1000):
  np.random.choice(N+1, N, replace = False)

eddjberry

\documentclass{article}
\usepackage{pdfpages}

\begin{document}

\includepdf[pages={30, 31, 32, 37}]{/path/to/file}

\end{document}

eddjberry

# R -------------------------
x = matrix(c(1,2,0, 4,3,7), ncol = 3, byrow = T)
x
#       [,1] [,2] [,3]
# [1,]    1    2    0
# [2,]    4    3    7

# row means (NB: R indexes from 1)
> apply(x, MARGIN = 1, mean)
# 1.000000 4.666667

eddjberry

# Dependencies ----------------------
import math 
import shap
import matplotlib.pyplot as plt

# shap_dependence_plot_grid ---------
def shap_dependence_plot_grid(cols, 
                              shap_values, 
                              X,    
                              interaction_index = None,