Telco Churn Analysis

Developed a churn prediction model using AllKNN with hyperparameter tuning, focused on minimizing false negatives. Achieved 93.7% recall, reducing potential high-risk churn losses by $18.8K and cutting total misclassification costs by $48.5K, outperforming benchmark models like XGBoost and Random Forest.

Proof points

  • 93.7% Recall
  • -$48.5K Costs

Technologies

Python, Scikit-learn, XGBoost, Pandas

Links

Problem

Telecom companies lose significant revenue from customer churn. Traditional models optimize for accuracy, missing high-risk churners.

Approach

Focused on recall optimization to catch high-risk customers. Used AllKNN with hyperparameter tuning, comparing against XGBoost and Random Forest.

Results

Achieved 93.7% recall, reducing high-risk churn losses by $18.8K and total misclassification costs by $48.5K compared to baseline models.