Customer Retention ML Model

Problem

Companies lose customers without knowing why. This model predicts churn early.

Dataset

10,000+ customer records including usage, billing, and demographics.

Approach

• Cleaned and preprocessed raw customer data
• Engineered behavioral features (usage frequency, billing trends)
• Built Logistic Regression baseline model
• Evaluated using confusion matrix and accuracy metrics

Key Insights

• Behavioral data is more predictive than demographic data
• Handling missing values significantly improved model accuracy
• Feature selection reduced overfitting