Extract hidden patterns in students' academic information to improve the curriculum by using data mining

Abstract

AbstractEducational data mining is an emerging exquisite field that has been successfully implemented in higher education. One of the best ways to increase the efficiency of this emerging phenomenon is to select efficient professors and effective teaching methods. This study is intended to show academic success factors to have better management in student curriculum, contextualizing the progress and to prevent unfavorable conditions for students. In this research, students of Shahrood University of Technology were studied. Initially, 3,765 samples of students' educational background were considered. Then, pre-processing was performed to make the data normalized. Next, several predictive models were developed using a supervised data mining approach. Next, five algorithms by the best result were selected. Comparing the results of algorithms applied to data, the two algorithms, radial basis function network and the Naïve Bayes with respectively value F-measure 0.929 and 0.942 showed more accurate results. Finally, the most effective feature was selected, the attributes ‘maximum semester’ and ‘the total number of units dropped’ were ranked an the most important, and the three attributes of ‘the basic courses average’, ‘the number of units of main courses’ and ‘the total average’, were placed in the next ranks.