Creditcard User Churn Prediction

Posted on 2025-09-2025

Tags: Python, Scikit-Learn, ML Classification, Data Cleaning, Precision Optimization

Problem:

A financial institution aimed to identify customers likely to stop using their credit card services. The business goal was to create a churn prediction model that would minimize false positives while helping the bank proactively reduce customer attrition.

Approach:

• Cleaned and explored customer account data including transaction counts, credit utilization, and tenure
• Engineered features to improve signal quality, including total transaction behavior and customer tenure categories
• Trained multiple classification models including:
  • Random Forest
  • Gradient Boosting
  • Logistic Regression (baseline)
• Focused model selection on maximizing precision to reduce cost of incorrectly flagging loyal customers

Outcome:

• Achieved 98% precision with the tuned classifier on the validation set
• Enabled the business to identify high-risk customers while avoiding over-targeting stable users
• Demonstrated the effectiveness of model-driven churn prevention in reducing customer loss and improving ROI on retention campaigns

Repo: GitHub - Credit Card Users Churn Prediction
Demo: (Add notebook viewer or screenshots if available)