Published Paper Details:

This In the digital economy, businesses increasingly rely on data-driven approaches to personalize campaigns and maximize customer engagement. Predicting customer response to marketing campaigns is a crucial challenge in modern business analytics, where organizations seek to optimize targeting and maximize return on investment. This study explores predictive modeling of customer response behavior using machine learning techniques Random Forest and XGBoost with SHAP on a real-world marketing dataset containing demographic, behavioral, and spending attributes. The models are evaluated using classification metrics, confusion matrices, and ROC-AUC scores, ensuring a comprehensive assessment of predictive performance. Results indicate that while Random Forest achieved perfect classification on the dataset, suggesting potential overfitting, the XGBoost model coupled with SHAP offered comparably high predictive performance with the added advantage of robustness and explainability. XGBoost effectively captured complex feature interactions, such as age-gender effects, making it more reliable for real-world marketing applications. The results highlight the potential of explainable artificial intelligence (XAI) in enhancing targeted marketing strategies, while also identifying future directions for scaling the approach with real-world, multi-channel customer data. Furthermore, correlation heatmap analysis revealed significant behavioral patterns, highlighting the role of spending behavior and demographics in influencing engagement. These findings provide actionable insights for businesses to design data-driven, personalized marketing strategies. The research emphasizes the dual importance of predictive accuracy and interpretability, ensuring models are not only effective but also transparent and usable for real-world decision-making.