Credit Card Churn Prediction: An Analytical and Model-Driven Study

Abstract

This study compares the performance of three machine learning models namely Logistic Regression, Naive Bayes, and K-Nearest Neighbors (KNN) in forecasting customer churn within financial institutions. Predicting customer churn is important for banks and the financial sector because once firms develop targeted interventions to improve customer satisfaction and loyalty, they can be kept for the long run. A set of performance metrics like accuracy, precision, recall, F1 score, and area under the curve were used in the analysis for the better comparison of the prediction capabilities of models.

Logistic Regression resulted as the one which performed best. It gained the highest possible accuracy of 89.6% and its recall rate was very good at 96.8%. Also, its AUC score shows high discriminative power at 0.91. Naive Bayes, gaining less accuracy by producing 86.7% yet showed good precision at 92.0%, and having good recall rates, 92.2% and thus forms another competitive selection. Its AUC score of 0.83 establishes its efficiency to differentiate churners from non-churners. KNN's accuracy was good with 89.2% along with excellent recall rate 97.6%. Moreover, AUC score at 0.86 enhances the reliability as well as performance in the role of a good prediction model for churners.

Promising results may indicate more research on such advanced techniques. More models that would improve the performance may be neural network models in this regard such as Artificial Neural Network, Feed Forward Neural Network, Multi-layer Perceptron in the sub-field of neural networks and the models applied to Deep Learning that may involve models such as Convolutional Neural Network, Recurrent Neural Networks, Long Short-Term Memory. These techniques are expected to further enhance predictive accuracy by capturing complex patterns in large datasets.

This would greatly enhance customer retention for banks and financial institutions in their implementation of machine learning models. The accurate prediction of churn will allow organizations to engage at-risk customers proactively, providing customized interventions to enhance satisfaction and loyalty. The results of this study further highlight the value of machine learning in changing customer relationship management and leading to long-term customer retention and organizational success.