Customer Loyalty Prediction of E-marketplaces Via Review Analysis

Abstract

Today, leveraging analytical CRM to maximize values for both customers and businesses is one the most important critical success factors. Predicting customer loyalty enables businesses to differentiate among customers for conducting relationship marketing and implementing effective customer extension tactics. In this paper, we analyze customers' reviews on the Digikala e-marketplace to predict their loyalty. We employ NLP, deep learning, and conventional machine learning methods and evaluate the results to find the best prediction model. Two experiments are conducted to evaluate the results: binary and 3-class loyalty prediction. In the binary setting, the Random Forest and Naïve Bayes algorithms outperformed the other tested classification methods and achieved an accuracy of 89%. In the 3- class setting, the Random Forest classification method achieved the best performance among all other machine learning algorithms with an accuracy of 67%. The evaluation results imply that businesses could benefit from using the Random Forest classification algorithm to predict customer loyalty through review analysis successfully. © 2023 IEEE.