A Machine Learning Approach to Predictive Modelling of Student Performance.

Area of Science:

• Educational Data Mining
• Machine Learning in Education

Background:

• Student academic performance is influenced by diverse factors including background, habits, and social activities.
• Identifying these factors is crucial for developing targeted interventions to improve student outcomes.

Purpose of the Study:

• To apply a data mining approach to identify significant factors influencing student performance.
• To predict student academic success using machine learning models.

Main Methods:

• Two datasets from Portuguese secondary schools were merged and pre-processed, with features normalized using linear scaling.
• Boruta feature selection was employed to identify relevant predictors, followed by classification using Support Vector Machine (SVM), Naïve Bayes (NB), and Multilayer Perceptron (MLP) models.
• Hyperparameter tuning was performed using GridSearchCV, with model performance evaluated using accuracy, precision, recall, and F1-Score.

Main Results:

• The Support Vector Machine (SVM) model achieved the highest accuracy in predicting student performance across different feature sets and classification levels.
• Specifically, SVM demonstrated strong predictive power, with accuracy scores reaching up to 91% for binary level classification based on historical grades.
• The study confirmed that a student's history of grades is a highly significant factor influencing their academic performance.

Conclusions:

• The history of grades is a critical predictor of student academic performance.
• Data mining techniques, particularly SVM, are effective tools for identifying key factors and predicting student success.
• Findings can inform educational strategies aimed at improving student outcomes.