Exploring Hybrid Models for Predicting LET Exam Takers' Performance: A Comparative Study of Decision Tree-ANN and Other Hybrid Approaches

This study presents a comparative analysis of hybrid machine learning models for predicting performance in the Licensure Examination for Teachers (LET). The study evaluates three hybrid approaches, which are the Artificial Neural Network with Decision Tree (ANN-DT), Support Vector Machine with Random Forest (SVM-RF), and Random Tree with Multilayer Perceptron (RT-MLP). A dataset composed of academic and demographic features of LET takers was used. This study assessed the models based on accuracy, precision, recall, F1-score, and Receiver Operating Characteristic (ROC) curve analysis. Among the models tested, the ANN-DT hybrid achieved the highest performance with an accuracy of 93.70%, balanced precision and recall, as well as gaining the highest area under the curve (ROC), demonstrating a robust classification capability. The SVM-RF model followed with good performance metrics, while the RT-MLP model showed lower predictive effectiveness. The study highlights the potential of using a hybrid model in optimizing performance prediction, which may contribute to the development of a knowledge-based system in identifying atrisk students to provide a timely academic intervention.