In this repo, a bank customer churn prediction model was developed. Various EDA and preprocessing steps were carried out, Number of ML algorithms tested and the better performing algorithm was selected for modelling and the hyper parmaeters tuned. Finally testing the model on unseen data to evaluate it.
The Best model is LGBMClassifier model with a learning rate of 0.05, max depth of 5, and 100 estimators achieved an accuracy of 0.8670. It showed a precision of 0.7571 and an F1 score of 0.5844. The model's ROC-AUC score was 0.8741, indicating good performance in distinguishing between positive and negative cases. Overall, the model demonstrated promising results, although further improvements can be made. Trialing with ensembles may result in better results