Optimizing Customer Relationship Management (CRM) Based on Machine Learning-Based Churn Prediction Models

In this era of changing marketplaces and intense rivalry among companies, customer relationship management (CRM) and data-driven marketing have become critical. Retaining current clients is one of the most crucial aspects of CRM. In order to forecast customer churn, this research makes use of ML algorithms. A classifier-based model for churn prediction using customer data is presented in this study. These trials make use of four different ML classifiers. The first step is to forecast the customer's churn status using these ML classifiers. Use the ADASYN to zero in on the problems with unbalanced datasets. This is to ascertain an efficacy of the model by measuring its F1-score, precision, recall, and accuracy. XGBoost outperforms all other models with the highest accuracy (96.71%), precision (87.60%), recall (92.62%), and F1-score (90.04%), demonstrating its effectiveness in predicting customer churn. While RF also performs well, CNN struggles to match traditional models, and LR shows a trade-off between recall and precision. The work adds to the body of literature on churn prediction by highlighting the integration of DL with conventional ML techniques and proving the better performance of XGBoost.