abhinav-neil / customer-churn

Analyze and predict bank customer churn using various classification algorithms

Customer Churn Prediction and Analysis

Introduction

We perform customer churn prediction on a bank customer dataset. The task is to predict whether a customer will exit the bank or not, based on his/her characteristics.

Data

The dataset contains information on a bank's customers, including various features such as age, geography, gender, income, balance etc., as well as whether or not they exited the bank. Source

Stages

1. EDA
   • Plot distributions of customers based on various features such as age, geography, income etc.
   • Plot conditional distributions of these features for exiting/retained customers to derive insights
2. Preprocessing and Feature engineering
   • Convert categorical features to one-hot representation for classification
   • Split data into train and test sets for features and target
3. Logistic Regression Model
   • Training: Fit logistic regression model to train data and make predictions on test data
   • Evaluation: Plot confusion matrix and ROC curve and compute area under ROC curve
4. Decision Tree Model
   • Training: Fit decision tree model to train data and make predictions on test data
   • Plot decision tree for visualization of decision nodes
   • Evaluation: Plot confusion matrix and ROC curve and compute area under ROC curve
   • Causal inference: Analyze important features for trained model
5. Random Forest Model
   • Training: Fit random forest model to train data and make predictions on test data
   • Evaluation: Plot confusion matrix and ROC curve and compute area under ROC curve
   • Causal inference: Analyze important features for trained model