SMOTE and Hyperparameter Optimization: A Dual Machine Learning Strategy for Enhancing Coupon Recommendation in Vehicular Contexts

This study centers on enhancing the accurateness of coupon recommendation algorithms within the framework of in-vehicle settings. The In-Vehicle Coupon Recommendation dataset from the UCI Repository was utilized in this study. A range of machine learning algorithms, such as Random Forest, Gradient Boosting, Extra Trees, LightGBM, XGBoost, AdaBoost, Stochastic Gradient Descent (SGD), and CatBoost, were employed for analysis. Among the several algorithms considered, CatBoost exhibited notable performance, with a baseline accuracy of 74%. In order to improve the efficacy of the model, we utilized the “Synthetic Minority Over-sampling Technique” (SMOTE) to mitigate the issue of class imbalance. This was subsequently followed by a thorough hyperparameter adjustment of the CatBoost algorithm. The improved model demonstrated a significant upgrade, with an accuracy rate of 77%. Additionally, there were noteworthy improvements in precision, recall, and F1 score. The AUC, which represents the Area Under the Receiver Operating Characteristic Curve, saw a notable improvement to 0.82. This signifies a noteworthy enhancement in the capacity to differentiate between positive and negative cases. The results emphasize the efficacy of CatBoost in the context of recommending coupons within vehicles, while also drawing attention to the influence of preprocessing methods and hyperparameter adjustment in enhancing the model's overall performance. This study is a valuable contribution to the domain of personalized advertising in the context of vehicles, providing valuable insights for the advancement of coupon recommendation systems that are more precise and dependable.