Educational data mining for predicting the academic performance of university students

Progress in the field of data mining makes it possible to use educational data to improve the quality of educational processes. This article examines various methods of analyzing student achievement data. The focus is on two aspects: first, predicting students' academic achievements at the end of a four-year undergraduate curriculum; second, examining typical student progressions and combining them with the prediction results. Approximately 10 classification algorithms were used in the prediction process. An approach to improving the performance of classification methods is proposed where classifier attributes are selected during their training. Two important groups of students were identified: low-achieving and high-achieving students. The results show that by focusing on a small number of courses that are indicators of particularly good or poor performance, it is possible to prevent and support low-achieving students in a timely manner, and to provide advice and opportunities to high-achieving students.