PERFORMANCE EVALUATION OF MACHINE LEARNING ALGORITHMS FOR PREDICTING STUDENT ACADEMIC OUTCOMES

Abstract

The growing access to educational information has provided substantial opportunities to use machine learning methods to forecast student performance. The research is an evaluation of the effectiveness of various monitored machine learning algorithms to predict academic success of students based on demographic, academic, and behavioral data. The type of quantitative experimental research design was adopted, and a structured dataset of the undergraduate students record was used. Data normalization, feature encoding, and dimensionality reduction methods were employed to preprocess the data, which does improve the predictive power of the study. An example of 10-fold cross-validation to implement and compare 6 algorithms, such as Logistic Regression, Decision Tree, Random Forest, Support Vector machine (SVM), XGBoost, and Artificial Neural Network (ANN) was performed. Accuracy, Precision, Recall, F1-Score and Area Under the ROC Curve (AUC-ROC) were used to evaluate model performance, and the results indicate that ensemble and boosting models are better than traditional models with XGBoost having the highest predictive accuracy (92%) and AUC-ROC (94%). ANNs and RFs also performed highly, which proves the effectiveness of nonlinear modeling methods in the educational data mining. The feature importance analysis showed that the past semester GPA, attendance percentage, and continuous assessment scores are the strongest predictors of academic outcomes. The results demonstrate the promise of machine learning-based predictive systems as valid early warning systems to detect at-risk students and facilitate academic interventions based on the available data.This study also adds to the existing body of research on educational analytics in the sense that it presents a full comparative analysis framework of machine learning algorithms in student performance prediction.

Keywords : Machine Learning; Student Academic Performance; Educational Data Mining; Predictive Analytics; XGBoost; Artificial Neural Networks; Early Warning Systems; Learning Analytics