Ensemble Learning for Student Performance Assessment: Identifying and Analyzing Significant Affecting Factors in Higher Education

This study aims to evaluate and enhance student performance prediction through a comprehensive analysis of various models. The objective is to identify the most effective model by comparing existing methodologies and proposing an ensemble model. The methodology involves data collection from openly sourced datasets encompassing academic backgrounds, personal habits, family dynamics, and health statuses. Preprocessing addresses inconsistencies and missing values, followed by exploratory data analysis (EDA) to understand data distributions and correlations. Ensemble modelling integrates algorithms like Multilayer Perceptron, Decision Tree, Logistic Regression, Naive Bayes, Random Forest, KNN, and SVM to enhance prediction accuracy. Significant factors affecting student performance are determined through statistical tests and appended to the dataset for further analysis. The results present a comparative analysis of various models, including a two-layer ensemble learning technique, multiple algorithms, Random Forest, and the Proposed Ensemble Model. The Proposed Ensemble Model outperforms all others, achieving perfect scores across accuracy, precision, recall, and F-score (all 1.00). Significant factors influencing student performance include academic achievement metrics, study habits, and health status. These findings underscore the superiority of the Proposed Ensemble Model in accurately predicting student outcomes, highlighting its potential for practical application in educational settings.