E-Commerce Customer Churn Prediction By Gradient Boosted Trees

The amount of data stored daily is increasing at a specific rate. E-commerce services are one of the areas where new knowledge is gathered on a daily basis. Therefore, it seems necessary to use data mining techniques in this field. This article aims to gain insight into a data set provided by the most important online food ordering service in Tehran, Iran. Data analysis can assist in discovering the causes of customer churn and also employ information to keep possession of customers. Customer churn is a significant criterion for evaluating a growing business, so it is important for companies to anticipate dominance to retain their customers. The aim of this article is the prediction of customer churn using online properties and user behavior. Multiple experiments have been performed to compare the result of distinct data mining methods. The results prove that Gradient Boosted Trees is better at the accuracy of 86.90%, among other techniques.