Student Academic Performance Prediction using Educational Data Mining

The field of Educational Data Mining concentrates on prediction more often as compared to generating exact results for future purpose. In order to keep track of the changes occurring in curriculum patterns, a regular analysis is a must in the educational field. This paper presents the initial results from an Educational data mining research project implemented by collecting huge data sets from BMS College of Engineering. It is aimed at revealing the high potential of data mining applications for university management and preventing the students from the risks of getting a low-Grade Point Average (GPA). WEKA platform was used to perform analysis with various machine learning models for both the classification and regression problems. Various classification and regression algorithms were analyzed and tested for accuracy for a student's performance in subject wise analysis and GPA prediction respectively. For the former, ensemble method using voting was performed on the classification algorithms. They were then categorized in different groups and compared based on certain factors, the results of which are discussed in this paper. For the latter, the Random Forest Algorithm gave the best results using regression analysis.