Telecom Churn Classification Using Scatter Search and Random Forest Classifier

Abstract

The Telecom Churn Dataset is a valuable resource that contains customer-related information such as demographics, usage patterns, billing data, and customer service interactions. The proposed model is used to explore various factors that might impact churn behavior and apply machine learning techniques to create predictive models for proactive retention efforts. This study have employed different approaches and methods, including random forest, logistic regression, and decision trees, as well as feature selection techniques like Correlation-based Feature Selection (CFS), scatter search etc. The proposed methodology involves preprocessing the data using techniques like SMOTE to handle class imbalance, selecting relevant features through Scatter search, and applying classification algorithms like the Random Forest (RF) classifier. The proposed model using Scatter search and Random Forest classifier has best formed with an accuracy of 91.9%, precision and recall of 0.919. Thus the model is well suited for telephone churn classification with the optimized feature set selection and classification. © 2023 IEEE.