This study presents a SHAP-based model for detecting customer fake reviews using SMOTE and hybrid learning algorithms to address the growing challenge of fake online reviews that undermine consumer trust, purchasing decisions, and business reputation. The proposed detection model integrates SHapley Additive exPlanations (SHAP), the Synthetic Minority Over-sampling Technique (SMOTE), and a hybrid Logistic Regression and Random Forest (LR–RF) model. Logistic Regression provides transparency, Random Forest captures complex relationships, SMOTE resolves class imbalance, and SHAP enhances model interpretability. The model was implemented in Python using standard machine learning libraries and evaluated on Amazon, Yelp, and TripAdvisor datasets with textual, behavioral, and metadata features to ensure cross-platform robustness. The experimental result showed that the model achieved 92.3% accuracy for single-domain and 89.0% for cross-domain detection. The proposed system promotes trustworthy reviews, supports business integrity, ensures fair competition, and aligns with regulatory requirements for transparent, explainable Artificial Intelligence (AI).
Keywords: SHAP, Customer fake reviews detection, SMOTE, Logistic Regression, Random Forest, Hybrid Model.