Comparative Performance Analysis of Boosting Ensemble Learning Models for Optimizing Marketing Promotion Strategy Classification

This study evaluates the performance of four boosting-based ensemble learning algorithms—AdaBoost, Gradient Boosting, XGBoost, and CatBoost—for classifying marketing promotion strategies. Using the Marketing Promotion Campaign Uplift Modeling dataset from Kaggle, it investigates how well each algorithm handles complex and imbalanced customer data involving demographic attributes, behavioral variables, and historical responses to campaigns. Model performance is assessed using accuracy, precision, recall, F1-score, and Area Under the Curve (AUC). The results show that XGBoost achieves the best precision, while Gradient Boosting attains the highest AUC, indicating superior discrimination between positive and negative classes. CatBoost demonstrates stable performance, especially with categorical data, whereas AdaBoost performs well in recall but is more susceptible to false-positive predictions. Although all four algorithms perform well overall, class imbalance remains a key challenge. The findings provide guidance for marketing practitioners in selecting suitable algorithms and underscore the importance of data-balancing strategies to enhance predictive accuracy in data-driven marketing.