Costumer churn Prediction with Gradient Boosting Machine. A Comparative Study

Authors

  • GODIAL Khalid, MOUTACHAOUIK Hicham

Abstract

Acquiring and retaining customers are two main concerns of  today's  business world.  The rapid increase in the market in each company leads to an increase in the number of customers. As a result, businesses have understood the importance of retaining these customers. It has become imperative for these companies to predict the customer churn rate of these customers, as negligence could lead to lower profitability from a major point of view. Forecasting the latter makes it possible to identify the clients likely to leave the company. Telecommunication, banking and human resources management are sectors among many others that have struggled against the risk of losing its customer base. this study implemented a comparison of the most used classification algorithms in related work and the Gradient Boosting Machine algorithms, to predict the churn rate of customers in the Telecom, bank and Human Resource sectors. The experimental results show that the models based on the Gradient Boosting Machine like LightGBM and CatBoost gave the best scores of Accuracies, Auc and Training time plus a compromise between sensitivity and specificity compared to other models.

Published

2020-11-01

Issue

Section

Articles