Research on Customer Churn Model with Least Square Support Vector Machine

Abstract

China telecommunications market is becoming more competitive, the operators are facing the severe costumer churn problem, and how to predict and effectively reduce the costumer churn directly concerns the survival and development of every operator. Therefore, the least squares support vector machine(LS-SVM) algorithm is adopted to build customer churn model, mainly including data cleaning, normalization, building forecasting model, model prediction, etc. The case study shows that the customer churn prediction using the LS-LSV has high precision, small error and remarkable effect.