Customer Churn Prediction in E-Commerce Using Machine Learning and Deep Learning Techniques

Pratik Shantaram Wagaskar, Dr. Usha Shete

Abstract: Churn of customers can be seen as a key problem for e-commerce businesses due to their impact on company's revenues. In current conditions, the task of determining customers, who will not use any services, becomes relevant. In this context, the task of predicting customers churn by means of Machine Learning and Deep Learning algorithms is chosen as the primary goal for this work. In particular, the objective of this study consists in the identification and comparison of various predictive models and selection of one model, which would demonstrate better performance. Customer dataset was selected for this work, and preliminary data preparation, namely missing values handling, data encoding, and feature scaling, were done. Then, some models, such as Logistic Regression, Decision Tree, Random Forest, and Artificial Neural Network, were implemented and evaluated by accuracy, precision, recall, and F1-score measures [1]- [3], [5]. Consequently, it became clear that ensemble and deep learning approaches work better in churn problem than simple models, and, in turn, the best results in terms of stability and accuracy were demonstrated by Random Forest and Neural Network, correspondingly.

Keywords: Customer Churn, E-commerce, Machine Learning, Deep Learning, Predictive Analysis, Customer Retention, Artificial Intelligence