An improved Ant Colony Optimization to Uncover Customer Characteristics for Churn Prediction

Document Type : Original Article

Authors

1 Department of Electrical & Electronics Engineering, Jazan University, Jazan, Saudi Arabia

2 Department of Electrical and Electronics Engineering , Jazan University, Jazan, 45142, Saudi Arabia

3 Department of Computer Science, Jazan University, Jazan, 45142, Saudi Arabia

Abstract

Feature selection (FS) is integral to machine learning applications for selecting a subset of salient features to improve performance and reduce computational time (CT). An enhanced Ant Colony Optimization (ACO) is presented for improving customer churn prediction models in the telecom industry. A weighted function of churn prediction performance and the number of optimal customer characteristics is optimized using ACO. The effect of hyper-parameters, like the pheromone value, heuristic information, pheromone decay factor, and the number of ants, on the optimization process is investigated. The optimization objective is measured by evaluating the prediction performance of selected features using the k-nearest neighbor classifier due to its simplicity and robustness. Comparative analysis using three different open-source customer churn prediction datasets showed that tunned ACO performs significantly better than other metaheuristic optimizers. The Friedman and Holms test validates improved performance by the tuned ACO. The selected optimal customer characteristics can be utilized to offer valuable insights for reducing customer churning.

Keywords

Main Subjects

Files

Share

How to cite

Statistics