Enhanced AI Classification Framework Using Hybrid Decision Trees and Ensemble Learning Techniques

In this paper, we provide an enhanced artificial intelligence (AI) classification strategy for educational achievement prediction. As models that are reusable, Decision Trees (DT) are subject to overfitting and poor generalization. To significantly reduce the number of dimensions, our model combines Principal Component Analysis (PCA) with collaborative learning methods, particularly Random Forest, AdaBoost, and Gradient Boosting.A Python-based experimental framework is implemented using data on actual pupil achievement. The use of PCA and scaling of features greatly increased generalisation achievement, decreased model complexity, and increased training efficiency. Furthermore, GridSearchCV was employed for systematic hyperparameter optimization, leading to noticeable improvements in the performance of the proposed hybrid classifiers. Among the evaluated models, the With 95% accuracy, 89% recall, and a 92% F1-score, the random forest classifier produced the best results. In terms of classification accuracy, the suggested method performs 17.3% better than the baseline Decision Tree model. These results show that the creation of scalable, dependable, and resilient AI-driven instructional decision-making systems can be aided by combining ensemble learning methods with PCA. Future research will solve class imbalance and expand the framework to multi-class categorisation contexts. This study helps provide a modular, comprehensible, and useful approach for academic decision-making.