lucdangelis

library(tidyr)
library(reshape2)
library(ggplot2)
library(corrplot)

#split numeric and categorical variables
is.fact <- sapply(df1, is.factor)
df1_cat <- df1[,is.fact] %>%
            mutate(count = 1) %>%
            gather

lucdangelis

df <- subset(df, select = -c( EmployeeCount, EmployeeID, Over18, StandardHours))
df1 <- df[complete.cases(df),]

lucdangelis

library(dplyr)

employee_survey_data <- read.csv('Downloads/HR Analytics/employee_survey_data.csv')
general_data <- read.csv('Downloads/HR Analytics/general_data.csv')
manager_survey_data <- read.csv('Downloads/HR Analytics/manager_survey_data.csv')

df <- general_data %>%
    left_join(employee_survey_data, by = "EmployeeID") %>%
    left_join(manager_survey_data, by = "EmployeeID")
str(df)

lucdangelis

#Plot historical probability of being alive
customer_id = ..
days_since_birth = ....
sp_trans = transaction_data.loc[transaction_data['customer_id'] == customer_id]
plot_history_alive(bgf, days_since_birth, sp_trans, 'date')

lucdangelis

#Calibration and Holdouts periods for data split
date_start = ['2013-04-01','2014-04-01','2015-04-01','2016-04-01']
calibration_period_end = ['2015-03-31','2016-03-31','2017-03-31','2018-03-31']
date_end = ['2016-03-31','2017-03-31','2018-03-31','2019-03-31']

#Arrays where to store the results of cross validation
accuracies_1y = []
holdouts_1y = []
predictions_1y = []

lucdangelis

plot_frequency_recency_matrix(bgf)
plot_probability_alive_matrix(bgf)

lucdangelis

cal_hold = calibration_and_holdout_data(trans_dataset, 
                                   'customer_id', 
                                   'date',
                                   calibration_period_end='2018-04-30', #2 years calibration
                                   observation_period_end='2019-04-30', #1 year holdout
                                   freq = 'M')   
cal_hold.head()

bgf = BetaGeoFitter()
bgf.fit(cal_hold['frequency'], cal_hold['recency'], cal_hold['T'])

lucdangelis

plot_period_transactions(bgf)

lucdangelis

import pandas as pd
import numpy as np
from lifetimes.utils import *
from lifetimes import BetaGeoFitter
from lifetimes.plotting import plot_probability_alive_matrix, plot_frequency_recency_matrix
from lifetimes.generate_data import beta_geometric_nbd_model
import matplotlib.pyplot as plt
from lifetimes.plotting import plot_calibration_purchases_vs_holdout_purchases, plot_period_transactions,plot_history_alive

bgf = BetaGeoFitter()

lucdangelis

beta_a = 2
beta_b = 3

for customer in range(0, 10):
  p_arr = []
  beta = np.random.beta(a=beta_a, b=beta_b)
  for transaction in range(1,10): 
    proba_inactive = beta*(1-beta)**(transaction-1)
    p_arr.append(proba_inactive)
  p_arr = np.array(p_arr)