Author(s) :   Dr. K. Nagamani, Dr. G. Ravi Kumar

Abstract : In every industry, including banking, e-commerce, and telecommunications, customer attrition is a serious problem. Organizations can proactively create retention strategies and minimize revenue loss by accurately forecasting attrition. Using the IBM Telco Customer Churn dataset as a case study, this paper introduces a generalizable machine learning approach for churn prediction in the telecom sector. Several machine learning techniques are used and assessed, including Random Forest, XGBoost, Multi-Layer Perceptron (MLP), and Logistic Regression. The feature selection method used is Recursive Feature Elimination (RFE), and the model's performance is evaluated both before and after dimensionality reduction based on RFE. Results from experiments show that while RFE increases interpretability and efficiency with no compromise in accuracy and recall, ensemble models (Random Forest and XGBoost) perform better in terms of prediction. The system offers a basis for data-driven churn management and may be tailored to various areas.

DOI : 10.61161/ijarcsms.v11i12.4

Pages : 24-29

*Authors are invited to submit papers through E-mail at editor.ijarcsms@gmail.com