An Explainable AI-Based Recommendation System for E-Commerce Platforms

Abstract

The widespread expansion of online retail has transformed recommender systems into an essential component of modern e-commerce platforms by guiding customers toward products that meet with their interests and purchasing behavior. Recent developments in machine learning have greatly enhanced the accuracy of recommendation models; however, several practical challenges still remain, many existing solutions provide little explanation of how individual recommendations are generated. This lack of interpretability can reduce user confidence and restrict the adoption of AIbased recommendation models in applications where transparent decision-making is required. To address this challenge, the present study introduces an Explainable Artificial Intelligence (XAI)-based recommendation method that brings together feature engineering, the Synthetic Minority Oversampling Technique(SMOTE), Extreme Gradient Boosting(XGBoost), and SHapley Additive exPlanations(SHAP). The workflow begins by preprocessing user–product interaction data and constructing informative features, including user activity, product popularity, and price buckets. The class imbalance problem is then handled using SMOTE before training an XGBoost classifier to estimate recommendation probabilities. To make the prediction process easier to understand, SHAP explains the contribution of each feature to individual recommendation outcomes at both the global and local levels. Experimental evaluation demonstrates that the proposed approach achieves strong predictive performance in terms of Accuracy, Precision, Recall, F1-score, and ROCAUC while maintaining a high level of model interpretability. By combining reliable prediction with meaningful explanations, the proposed recommendation approach strengthens user trust and offers a practical solution for deployment in intelligent e-commerce environments.