Machine Learning Techniques for Determining Students' Academic Performance: A Sustainable Development Case for Engineering Education

Abstract

This research paper presents the approach of machine learning analysis techniques on education. For the large dataset, data mining techniques are used to extract hidden information and create insight. We hypothesized that the prediction algorithm and dimensional reduction algorithm could be used on an educational dataset to extract the hidden information and analyze the information to create insight. Machine learning algorithms can be used to predict student academic performance. Since some of the features in our dataset are correlated so, before applying the prediction algorithm, we applied the dimensional reduction algorithm to reduce the dimension of our dataset and extract the important features. For the prediction analysis, we used three supervised machine learning algorithms, namely K-Nearest Neighbors (KNN), Decision Tree, and Logistic Regression. Before we applied these machine learning algorithms, we applied the dimension reduction algorithm for the feature extraction purpose using two algorithms, namely Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). We compared the performance of these machine learning algorithms. For the student academic performance, their final Examination result was taken as the target value, which was predicted by using the above-mentioned supervised algorithm. Our work shows that the dimensional reduction algorithm, followed by the prediction algorithm, achieved the acceptable prediction accuracy for determining student academic performance. Our result also highlights the advantage of employing machine learning techniques on educational data and explains how it helps to provide engineering education insight for the sustainable development of Engineering education as a whole.