Research Review on Intelligent Decision Systems Based on the Integration of Decision Trees and Ensemble Learning

Abstract

This paper summarizes the latest research progress of combining decision trees with ensemble learning in intelligent decision-making systems. The rules of the decision trees are clear, the reasons are simple, and it is easy to meet the characteristics of categorical and continuous at the same time. Therefore, financial risk control, medical diagnosis and industrial forecasting can always improve the stability, generalization ability and strength of resisting noisy data by combining multiple base learner. This study systematically investigates the optimization algorithm of decision trees, AdaBoost/Gradient Boosting and other enhancement methods, as well as the ensemble framework including Bagging, Boosting and Stacking. We pay attention to the application examples that use them together, such as software defect prediction, fault classification, breast cancer detection, SQL injection detection and IoT intrusion prevention. Key challenges in current research are identified: inadequate interpretability and visualization capabilities, weak defenses against adversarial samples and distribution drift, and immature online learning and lightweight frameworks. Finally, future research should prioritize aligning interpretability and robustness as co-optimization objectives, developing unified defense mechanisms, integrating automated feature engineering and lightweight ensemble models, and advancing adaptive updates for both online and edge computing scenarios. These discussions aim to establish a clearer research framework and direction for integrating decision trees with ensemble learning in intelligent systems.