Integrative Ensemble Learning Algorithm for Predicting Students’ Performance

Abstract

Listen

Translate article

Objectives: To create a stable student performance prediction model utilizing ensemble learning methods. Methods: The study uses boosting techniques such as CatBoost, Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LightGBM) as simple classifiers, which are then combined into a composite classifier to improve predictive accuracy. During the training phase, a 5-level hyperparameter optimization for the basic classifiers is performed using ETLBO Optimization IELA's distinguishing feature is its Stacking ensemble method, which functions as an ensemble technique, combining the expected outcomes of the simple classifiers to build the final prediction model ETLBO algorithm is applied for hyper parameter optimization to discover the best hyperparameters of 28 features from 33 features in the base classifiers to yield the better result in experiment. This approach improves prediction efficiency and accuracy by combining the capabilities of separate models using boosting algorithms and stacking-based ensemble techniques. Findings: Student achievement Dataset in secondary education for Mathematics are taken from Public repository is used in this research work. It comprises features such as student grades, demographic information, social aspects, and school-related data. The working of boosting and stacking approaches with the classifier in the dataset consisting of LightGBM and CatGB are performed higher than those of either classifier when used separately without the ensemble technique. Additionally, CatGB operates substantially better when combined with XGBoost. With an accuracy of 86.43% and an F-score of 84.98%, the composite classifier that combines the three simple classifiers achieves the highest gain. Novelty: This research is unusual to combine the boosting and stacking ensemble techniques, which improves prediction accuracy and efficiency when compared to previous models for predicting student performance. Keywords: Educational Data Mining, Ensemble Algorithm, Boosting Algorithm, Machine Learning Algorithm, Base Classifier, Composite Classifier