Development of a model for estimating and classifying the educational performance of undergraduate students using a combination of multilayer neural networks (Case study: Qom University)

Students’ growth and development have always been considered important by the education system as they are the future assets of their country. Although many students can flourish their talents and creativity, we are faced with a large number of students each year whose talents are squandered leading them to a position far from success. Today, with the more students entering in different academic levels and the variety of study disciplines, the necessity to properly guide students is felt more than ever. For this purpose, educational data mining has received special attention from the educational system’s officials in recent years. So far, various classification methods and techniques in data mining and machine learning have been used to predict student performance. However, these individual classifiers have limitations such as complexity and instability for predicting performance in the education process. To tackle this problem, ensemble classification has been proposed as a new and efficient method. Ensemble classification systems combine the results of several individual classifiers to provide a model with better performance. In this paper, a new Ensemble classification system is presented using multilayer neural networks and SOM clustering in order to estimate and classify the grade point average of undergraduate students. In addition, we used averaging and majority voting as combination methods for aggregating the results of individual classifiers. Evaluation results on real university data show that our proposed ensemble system provides better accuracy and performance compared to prevalent individual classification methods. Also, the proposed ensemble system obviously outperforms other popular ensemble methods in classifying students’ GPA.