Automatic detection of students at risk of dropping out of school using mRMR and Late Fusion

Different Educational Data Mining (EDM) techniques have been utilized to detect the risk of school dropouts over the previous 10 years, and the issue remains to improve the performance and interpretability of early detection models. Based on a metric design and classifier fusion technique, this research proposes an automatic dropout risk detection method. The goal of this study is to properly pick the characteristics that have the greatest impact on the probability of dropping out of school and use them as inputs to a classifier. Processing, modeling and evaluation were the three steps in the methodological design. The findings show that (1) the minimal-optimal method (mRMR) allows for a robust selection of the student’s characteristics with good discriminant capacity, (2) the transformation based on metrics aids in the enhancement of individual characteristics; and (3) the Model accuracy increased by 10% when classifier fusion was used instead of a single classifier.