Classification Trees

Objective

Predict attrition using the credit card data, using the classification algorithm from the rpart package.

Data Prep

Created a new binary variable cheater by classifying observations with 0 affairs as 0, and with one or more affairs as 1. Split the data (601 observations), into training and validation data sets (70%/30% , 6,097/4,030)

The Tree

Two Interpretations from the Tree

The most likely customers to leave are:
- Total Transaction Count (Last 12 months) < 55
- Total Revolving Balance on the Credit Card < 614
- Change in Transaction Count (Q4 over Q1) >= 0.65
- Total no. of products held by the customer < 3
Unlikely customers to leave are:
- Total Transaction Count (Last 12 months) >= 55
- Total Transaction Amount (Last 12 months) < 5271

Variable Importance

Interesting how Age is perceived by many as a predictor for customers leaving. However, the data shows that it is not a good predictor.

Model Accuracy

Using the validation data, it is observed that the model is able to:

Predict correctly 58.4% of customers leaving
Predict correctly 98.6% of customers staying
Misclassified faithful customers as leaving 1.4% of customers that were not going to leave

## 
##  
##    Cell Contents
## |-------------------------|
## |                       N |
## | Chi-square contribution |
## |           N / Row Total |
## |           N / Col Total |
## |         N / Table Total |
## |-------------------------|
## 
##  
## Total Observations in Table:  1063 
## 
##  
##                                | validation_tree$Attrition_predicted 
## validation_tree$Attrition_Flag | Attrited Customer | Existing Customer |         Row Total | 
## -------------------------------|-------------------|-------------------|-------------------|
##              Attrited Customer |                80 |                57 |               137 | 
##                                |           385.947 |            37.003 |                   | 
##                                |             0.584 |             0.416 |             0.129 | 
##                                |             0.860 |             0.059 |                   | 
##                                |             0.075 |             0.054 |                   | 
## -------------------------------|-------------------|-------------------|-------------------|
##              Existing Customer |                13 |               913 |               926 | 
##                                |            57.100 |             5.475 |                   | 
##                                |             0.014 |             0.986 |             0.871 | 
##                                |             0.140 |             0.941 |                   | 
##                                |             0.012 |             0.859 |                   | 
## -------------------------------|-------------------|-------------------|-------------------|
##                   Column Total |                93 |               970 |              1063 | 
##                                |             0.087 |             0.913 |                   | 
## -------------------------------|-------------------|-------------------|-------------------|
## 
##  
## Statistics for All Table Factors
## 
## 
## Pearson's Chi-squared test 
## ------------------------------------------------------------
## Chi^2 =  485.525     d.f. =  1     p =  1.341304e-107 
## 
## Pearson's Chi-squared test with Yates' continuity correction 
## ------------------------------------------------------------
## Chi^2 =  478.4127     d.f. =  1     p =  4.733108e-106 
## 
##