School Students' Performance Predication Using Data Mining Classification

Abstract

Educational management information systems generate huge amounts of data which hide a very useful knowledge. The techniques and methods used to discover the knowledge from students data are known as Educational Data Mining (EDM). The main objective of EDM is to improve students and teachers performance. Many researchers analysed students' behaviour to obtain useful knowledge that can help educators in planning for improving students' performance. There are two approaches which can be used to discover knowledge; by statistical methods and by DM techniques such as classification. This paper proposes a students' performance prediction model based on DM classification algorithms (Naïve Bayes, Decision Tree and K-NN). The dataset was collected from a preparatory male schoolin Gaza strip, includes over 1100 records. Obtained results show that Decision Tree gives the best results. Moreover, the results indicates that social case has little impact on the students' performance, while the academic features such as previous year and first term results have more impacts on the performance. These results can be used in improving students' performance by predication their retention early to minimize students' failure.