Data Fusion and Predictive Modeling For Academic Performance Assessment: A Case Study On Grade Variation

Education and academic success are of great importance in today's society, particularly for the professional, economic, and personal future of young people. With this in mind, it is of great interest and vital importance for educational institutions to be able to predict the variation in students' grades, especially students at risk of failing, since teaching methods can be changed, and corrective measures and intervention strategies can be applied to support underperforming students, taking their needs into account. The introduction of Artificial Intelligence and Data Fusion techniques in this area could be very interesting and could improve efficiency in detecting students at risk of school failure. This paper aims to develop a case study in which algorithms and procedures are developed to merge and integrate data and create predictive models. It is proposed to develop predictive models to predict variations in the grades of middle school students in mathematics from the first to the second term, by implementing an early fusion technique. In view of the best candidate models obtained, it was proven that data fusion performs well in creating predictive models for predicting grade variations and that both fusion techniques tested are competent, seemingly improving prediction results compared to models created from separate datasets.