Improved Cuckoo Harmony Search Optimization and Enhanced Support Vector Machine Algorithm for Churn Prediction on Telecom Data

Abstract

Customer Relationship Management (CRM) aims at the growth of commercial, long period association with key consumers and stakeholders. The main issue in CRM in the telecom industry is churning, where the customers could shift to a competitor with much ease. Research in the Telecom industry specified that it is an expensive procedure for the business, because the retention of unsatisfied customers becomes more complicated than attracting the new customers. This scenario operates practical research over customer churn to forecast and enhance the usage of classifiers to improve a customer churn prediction. In this research, the improved Cuckoo Harmony Search (ICHS) optimization algorithm with Enhanced Support Vector Machine (ESVM) algorithm is developed to improve the churn prediction performance significantly. This research compromise of four main phases such as pre-processing using Expectation Maximization (EM) clustering algorithm, customer behavior analysis using K-Nearest Neighbor (KNN) classifier, attribute selection using ICHS optimization algorithm, and customer churn prediction using ESVM classification algorithm. In the pre-processing step, EM algorithm is efficient in handling the missing values. Then K-Nearest Neighbor algorithm analyzes the customer behavior using K-closest similarity values by considering the four conditions such as “client displeasure (H1), switching expensive (H2), service utilization (H3), and client category (H4)”. In attribute selection, ICHS chose the best informative attributes. ESVM is used in this work as a basis prediction model for customer churn prediction by enhancing SVM using Particle Swarm Optimization (PSO) algorithm. Thus, the experimental results conclude that the proposed ICHS-ESVM method provides better churn prediction performance rather than the existing techniques.