An Attribute Weight Estimation Using Particle Swarm Optimization and Machine Learning Approaches for Customer Churn Prediction

One of the most challenging problems in the telecommunications industry is predicting customer churn (CCP). Decision-makers and business experts stressed that acquiring new clients is more expensive than maintaining current ones. From current churn data, business analysts must identify the causes for client turnover and behavior trends. This study uses PSO for feature selection and the four most powerful machine learning techniques to predict churn customers, including Decision Tree and K-Nearest Neighbor, Gradient Boosted Tree, and Naive Bayes. An experiment is conducted using two performance measures accuracy and precision. The proposed methodology initially employs classification algorithms to categorize churn customer data, with the Gradient Boosted Tree, Decision Tree, k-NN, and Naive Bayes performing well in accuracy, achieving 93 percent, 90 percent, 89 percent, and 89 percent, respectively. The experimental findings showed that the Gradient Boosted suggested methodology outperformed by obtaining an overall accuracy of 93 percent and precision of 87 percent, which shows the effectiveness of the proposed method.