期刊名称：International Journal of Computer Science and Network Security
印刷版ISSN：1738-7906
出版年度：2019
卷号：19
期号：9
页码：103-112
出版社：International Journal of Computer Science and Network Security
摘要：Student profile detection and performance prediction are both high-potential areas in the educational data mining domain. To meet these goals, multi-criteria analysis and machine learning techniques are employed to help with decision making when it comes to student failure and extracting useful, hidden and relevant information about students. In this paper, we propose a multilevel hybrid system for student performance prediction. This work combines two approaches: multi-criteria analysis via the method AHP (Analytic Hierarchical Process) and classification and multi-level prediction using machine learning techniques. Different classification techniques were compared, such as SVM, NB and DT, with the last one performing the best. An analysis using association rules was also conducted in order to detect the different hidden relationships between the scores obtained and the modules. To obtain a high performance in students’ failure prediction, we successfully aggregated machine learning methods with feature selection and parameter optimization process. The results show that the student performance prediction is efficiently done and sufficient performance is obtained. Hence, our system is able to identify at-risk students, assess the adequacy of the courses or modules, and help tailor interventions to improve on student success.
关键词：Educational data mining;prediction;profile analysis;student failure
