Construction of Churn Prediction Model Using Human Voice Emotions Features Based on Bayesian Belief Network

Abstract

Predicting customer churn to retain existing customers is a hot topic both in the world of academia and business today. One of them is research the prediction of churn based on customer emotions. Emotion is an important catalyst that affects customers in the process of purchasing services, customer satisfaction in goods and services products, and assessing the level of customer loyalty to companies in the future. Bayesian Belief Network (BBN) will be used in the construction of a churn prediction model that is based on four types of happy, sad, angry, and fear emotions. The results showed that the utilization of human emotional voice classification as a variable in churn prediction can provide predictive results on the Bayesian Belief Network with a churn value of 60% and not churn of 40%.