APPLICATION OF NEURAL NETWORKS TO PREDICT THE FUTURE PROSPECTIVE STUDENTS’ EDUCATIONAL ACHIEVEMENT AT THE UNIVERSITY

Abstract

The issues of improving the quality of education at the university by developing a recommendation to the applicant for admission to a university in a suitable field of study are considered. Predicting the student's academic performance for the chosen field of study reduces the risk of choosing the wrong direction, and, as a result, increases the level of mastery of appropriate competencies when studying at a university. The successful experience of foreign universities using machine learning methods to study the quality of students’ education in universities is considered. For forecasting, the authors use a neural network model describing the relationship between predictors and a criterion variable. Indicators reflecting the quality of the applicant’s education in secondary school were chosen as predictors. Objective data of indicators acting as predictors are considered: grades in school natural science disciplines, scores of the unified and main state exam in natural science disciplines, the average score of the certificate. As a criterion variable, the average assessment of the teachers of the graduating department was chosen, reflecting the level of mastering the core competencies. The results of the work of this model for real source data of students of the direction “Computer Science and Computer Engineering” of several years of study are presented. The effectiveness of the proposed forecasting model with a sufficient degree of accuracy for the available statistical data is shown. To check the adequacy of the proposed model, a multidimensional linear regression model is constructed. The main indicators of this model are described, their values are given. The conclusion is made about the success of the proposed forecasting model and suggests ways to further improve the accuracy of the forecast.