Using some popular ML algorithms to predict churning!

Loading the needed packages

library(tidyverse)
library(tidymodels)
library(themis)
library(vip)
library(xgboost)
library(rpart.plot)

Loading the datset

churn<-read_csv("churn.csv")
df<-churn %>% slice_head(n = 1000)
new_data<-churn %>% slice_tail(n = 10) %>% select(!churn)

1.XGBoost model

Predicting on new data

## # A tibble: 10 × 3
##    .pred_class .pred_churn .pred_no_churn
##    <fct>             <dbl>          <dbl>
##  1 no_churn       0.000384           1.00
##  2 no_churn       0.000384           1.00
##  3 no_churn       0.000384           1.00
##  4 no_churn       0.000384           1.00
##  5 no_churn       0.000384           1.00
##  6 no_churn       0.000384           1.00
##  7 no_churn       0.000384           1.00
##  8 no_churn       0.000173           1.00
##  9 no_churn       0.000384           1.00
## 10 no_churn       0.000360           1.00

2.Random Forest model

Predicting on new data

## # A tibble: 10 × 3
##    .pred_class .pred_churn .pred_no_churn
##    <fct>             <dbl>          <dbl>
##  1 no_churn         0.376           0.624
##  2 no_churn         0.376           0.624
##  3 no_churn         0.376           0.624
##  4 no_churn         0.376           0.624
##  5 no_churn         0.376           0.624
##  6 no_churn         0.376           0.624
##  7 no_churn         0.376           0.624
##  8 no_churn         0.335           0.665
##  9 no_churn         0.376           0.624
## 10 no_churn         0.0960          0.904

3.Decision Trees model

Predicting on new data

## # A tibble: 10 × 3
##    .pred_class .pred_churn .pred_no_churn
##    <fct>             <dbl>          <dbl>
##  1 no_churn         0.0148          0.985
##  2 no_churn         0.0148          0.985
##  3 no_churn         0.0148          0.985
##  4 no_churn         0.0148          0.985
##  5 no_churn         0.0148          0.985
##  6 no_churn         0.0148          0.985
##  7 no_churn         0.0148          0.985
##  8 no_churn         0.0148          0.985
##  9 no_churn         0.0148          0.985
## 10 no_churn         0.0148          0.985

4.Comparing results

5.Confusion matrices

6.ROC curves

7.Variable Importance

8.Plotting the Decision Tree