Data Mining in education: predicting student performance

The ability to predict student academic performance is valuable to any institution seeking to improve student achievement and motivation. Based on the predictions generated, students identified as being at risk for expulsion or failure can be supported in a more timely manner. This article discusses various classification models for predicting student performance using data collected from universities in Penza. The data include student enrollment data as well as activity data from the university electronic information and education environment (EIE). An important contribution of this study is the consideration for student heterogeneity in the construction of predictive models. This is based on the observation that students with different socio-demographic characteristics or modes of learning may exhibit different motivation to learn. Experiments confirmed the hypothesis that models trained using instances in student subgroups outperform models built using all data instances. In addition, the experiments showed that accounting for both enrollment and learning activity patterns helped to identify vulnerable students more accurately. Experimental results have demonstrated that no single method has superior performance in all aspects. The homegrown analytics platform Loginom was employed as a tool to create a predictive model.

1. Politov A.Yu., Akzhigitov R.R., Sudarikov K.A. Analiz modelei i instrumentov prediktivnoi analitiki dlya analiza obrazovatel'nykh dannykh. Innovatsii. Nauka. Obrazovanie. 2021;28:1055–1065. (In Russ.).

2. Salloum S.A., Elnagar A., Shaalan K., Alshurideh M. Mining in Educational Data: Review and Future Directions. Advances in Intelligent Systems and Computing. 2020;1153:92–102. DOI: 10.1007/978-3-030-44289-7_9.

3. Piskunov L.A. Analiz i prognozirovanie uspevaemosti studentov na osnove neironnoi seti. Vestnik nauki. 2019;3(6):402–405. (In Russ.).

4. Vasil'eva E.E., Kurushin D.S., Vlasov S.S. Early prediction of the grade point average of university students diploma: neural network approach. International Conference on Soft Computing and Measurements . 2019;1:366–369. (In Russ.).

5. Ozerova G.P., Pavlenko G.F. Prediction of student performance in blended learning utilizing learning analytics data. Science for Education Today. 2019;9(6):73–87. DOI: 10.15293/2658-6762.1906.05. (In Russ.).

6. Pevchenko S.S., Bluzhin V.A. Sravnitel'nyi analiz algoritmov neironnoi seti i derev'ev prinyatiya reshenii modeli intellektual'nogo analiza dannykh. Molodoi uchenyi = Young Scientist. 2016;132(28):148–154. (In Russ.).

7. Shukhman A.E., Parfenov D.I., Legashev L.V., Grishina L.S. Analysis and forecasting students’ academic performance using a digital educational environment. Vysshee obrazovanie v Rossii = Higher Education in Russia. 2021;30(8-9):125–133. DOI: 10.31992/0869-3617-2021-30-8-9-125-133. (In Russ.).

8. Gasevic D., Dawson S., Rogers T., Gasevic D. Learning analytics should not promote one size fits all: The effects of instructional conditions in predicting academic success. The Internet and Higher Education. 2016;28:68–84. DOI: 10.1016/j.iheduc.2015.10.002.

9. Akbarkhuzhaev S.A., Abdurakhmanova N.N. Sravnitel'nyi analiz metodov Naivnogo Baiesa i SVM algoritmov pri klassifikatsii tekstovykh dokumentov. Molodoi uchenyi = Young Scientist. 2019;267(29):8–10. (In Russ.).

10. Bulycheva P.A. Oshmarina O.E., Shadrina E.V. Identifying academically “unsuccessful” first-year university students: a case study of Higher School of Economics – Nizhni Novgorod. Vestnik Nizhegorodskogo universiteta im. N.I. Lobachevskogo. Seriya: Sotsial'nye nauki = Vestnik of Lobachevsky State University of Nizhni Novgorod. Series: Social Sciences. 2016;42(2):136–143. (In Russ.).