A Hybrid Machine Learning Model for Predicting Customer Churn in the Telecommunication Industry

The churn rate, being a major issue in the telecommunication industry today as it describes the degree at which customers desert a company and discard its services in one way or another. Over the past years, various changes evolved in the telecommunications industry like liberalization of the market, giving rise to serious competition and hence the evolution of new technologies and opportunities which directly and indirectly affect customer choice and preference for the services of a specific telecommunications company. In this study, a generalized K_LoRD hybrid model for predicting customer churn in the telecommunication industry was developed using K Nearest Neighbor, Logistic Regression, Random Forest and Decision Tree. A publicly available dataset from a telecom company which has a record of customer information and their churn was collected. The K_LoRD hybrid model is evaluated based on the following performance metrics: Accuracy and Receiver Operating Curve. The results showed that the hybrid model efficiently predicts customer churn with 91.85% prediction accuracy and 95.9% Area Under Curve. The experiments have shown that our hybrid prediction model is superior to ordinary K nearest Neighbor, Logistic Regression, Random Forest and Decision Trees.