Predicting Students' Performance by Using Data Mining Methods

Abstract

Abstract
 The corona pandemic disrupted the educational process, especially in universities that use traditional education. Universities were therefore obliged to move from traditional education to e-education without adequate preparations. The aim of this research is to analyze the students' performance in the two educational environments and predict the result of any of them in the future. The k-means algorithm, an important data mining method, was used to analyze the results of the fourth-stage classes of five consecutive years of students from one Iraqi university's scientific departments. Four of these years were traditional education, while the last was E-education to see whether the student's performance distribution is normal or abnormal. The results indicate a 100 percent of students’ success rate in e-education, while the upper limit is 70 percent for the previous years.
 Moreover, the average class rate increased to 75 percent compared to 62 in previous years. The decision tree has been built based on a dataset created from the collected data to predict the distribution of both traditional and e-education with a 2% error tolerance. The study shows that using the exact mechanism in e-education will give abnormal results. Therefore, the study recommends the need for good infrastructure, the preparation of efficient staff, increasing students’ skills, and appropriate software platforms for an accurate assessment of students’ performance.