A Two-Stage Ensemble Approach for Analysis of Optimizing Customer Churn with Lime Interpretability

Businesses must outbid suppliers to win over new clients in a very competitive industry. Since client retention directly affects a company’s revenue, it is a prominent topic for research. Businesses can prevent client attrition by proactively addressing it when it occurs. A two stage predictive ensemble method is proposed with Gaussian Naive Bayes, Perceptron, Gradient Boosting Classifier, Decision Tree, K-Nearest Neighbors, Logistic Regression, and XGBoost used in the first stage model classification, and then models like Adaboost, Bagging Classifier, Stacking, and Voting are used in the second stage to increase forecast accuracy for the top four models chosen in the first stage. The LIME Explainer algorithm also analyzes data instance adjustments and their effect on predictions, improving interpretability for “black box” models. This two-step approach addresses issues in crucial decision-making, by optimizing the final model’s predictive power and interpretability.