Formation of forecast models in assessing the quality of student training

The objective of the study is to create forecasting models for assessing students’ academic performance, attendance, and the risk of their expulsion from the university. The work is devoted to the development of forecasting models of students’ academic performance based on their current grades using data analysis methods. The Statistica program was used in the study. Regression analysis, binary choice models, and cluster analysis were used as tools. The main sources of data were the results of a survey of students of the economic, physical and mathematical, and electrical power engineering faculties (38 academic groups), as well as information on students expelled according to the deans’ offices of the EF and FMF of the Institute of Digital Technologies of the Mari State University. Three models were developed as part of the study: the first uses the “Number of debts” indicator as a dependent variable, the second - “Number of absences”, and the third - “Number of expulsions”. Regression statistical models, classifications, and clustering, as well as probabilistic models of discriminant and logit analysis, were built for each model. The article presents a methodology for constructing predictive assessments of students of the Institute of Digital Technologies of the Mari State University based on data from the electronic system “Student” and a survey of 290 students on 12 parameters. These parameters include indicators reflecting the ability to study (USE results, average grade point average), readiness for independent living (living in a dormitory), as well as the desire and ability to learn (current attendance and academic performance). Predictive models will allow for a timely assessment of the student’s work and the adoption of measures to correct the educational process of the university.