MVP.md

Classification of Customers Churn (MVP)

The purpose of the classification is to understand the influence of various factors on the bank customers who are closing their accounts (Churn).

Figure 1

features that has big correlation with exiting the bank

• Age (29%)
• Customers in Germany (17%)
• Female customers (11%)

Logistic regression before oversampling training and validation set metrics:

Training	Validation
Accuracy: 0.7869	Accuracy: 0.7894
Precision: 0.3864	Precision: 0.3757
Recall: 0.0520	Recall: 0.0539
F1: 0.0916	F1: 0.0943

Figure 2

The following figure is our baseline model before oversampling: Logistic Regression.

The score of our baseline was actually good with an accuracy of: 0.7869

Figure 3

The following figure is our data with oversampling

Logistic regression after oversampling training and validation set metrics:

Training	Validation
Accuracy: 0.6643	Accuracy: 0.6799
Precision: 0.6621	Precision: 0.6753
Recall: 0.6950	Recall: 0.7149
F1: 0.6950	F1: 0.6945

In conclusion after we did the oversampling, the accuracy has been reduced from 79% to 63%