Comparative Study of Prediction Models on High School Student Performance in Mathematics

Measuring students performance is a challenging task that can help students and teachers to keep track of progress of students performance. Mathematics is one of the basic pillars for many subjects, as well as the backbone of scientific undertaking in the age of information revolution. In order to improve the predicting results, several techniques have been investigated and compared to get the optimized prediction model. In this paper, we proposed comparative study of a statistical analysis techniques, machines learning (ML) algorithms, and one of deep learning architecture for predicting student performance in mathematics. A statistical technique structural equation modeling (SEM), five classes of machine learning algorithms, and a framework of deep learning, Deep Belief Network (DBN) were executed and compared. Two datasets namely, DS1, and DS2 of the same characteristics with different sizes were used. Through the three datasets, Random Forest (RF) was found to outperform others models in prediction the student performance.