An AI-Driven Classification System for Early Detection of Customer Churn in E-Commerce Platforms

Abstract

Customer churn is a key issue for e-commerce organizations since it lowers revenue and hampers long-term growth. E-commerce companies may reduce expensive customer attrition and implement proactive retention measures by predicting client churn. This study proposes an AI-driven classification system for early customer churn detection using the List online retail dataset. The proposed CNN has a higher ability in capturing complex and nonlinear patterns of customer behavior as opposed to the conventional machine learning methods. The results of the experiment display excellent performance, with 99.62% accuracy, 99.32% precision, 98.66% recall, and 98.78% F1-score and with the AUC of 0.9893. The findings demonstrate that the model is robust, has generalization and small misclassification with respect to the baseline models like Random Forest, Logistic Regression, AdaBoost and ANN-MLP. In addition to predictive strength, the paper highlights an importance of DL in extracting latent features representations and provides actionable information on customer retention strategies. The results determine CNNs as a strong and justified solution in detecting churn that provides academic value and practical applicability to the e-commerce industry.