Problem

A bank was experiencing a high rate of customer churn, which was impacting their revenue and growth. They needed to identify the factors that were causing customers to leave and develop a model that could predict which customers were at risk of churning.

The bank had a large dataset with information about customer demographics, account activity, and transaction history. However, the data was unstructured and needed to be preprocessed before it could be used for modeling.

Learn more about our approach here -https://github.com/RushikeshRC0387/Bank-customer-Churn-Model

Solution

We developed a machine learning model using Support Vector Machine (SVM) classifier to predict customer churn. We started by preprocessing the data, encoding categorical variables, scaling features, and balancing the dataset using random under sampling and over sampling techniques.

After training the model on the preprocessed data, we achieved an accuracy of 84%. To optimize the model further, we performed hyperparameter tuning with Grid Search and achieved a balanced accuracy of 80%.

Impact

Reduced customer churn rate.
Increase in revenue.

Team

Rushikesh Choudhari - Data Scientist
- Preprocessed the data, developed and trained the machine learning model
- Performed hyperparameter tuning with Grid Search to optimize the model
Rushikesh Choudhari - Business Analyst
- Analyzed the impact of customer churn on revenue and growth
- Developed a strategy to retain customers at risk of churning

Bank Reduces Churn with SVM Classifier Model