Introduction. The article examines the problem of assessing students' academic performance in the current situation. The purpose of the paper is to evaluate the influence of e-learning and some social and behavioral parameters on students' academic performance. Materials and Methods. The author employed the machine learning procedures in order to identify and assess the current problems of the educational system, students' behavior, and universities' policy. Methods of mathematical analysis and statistics as well as ensemble methods (gradient boosting and the random forest algorithms) were used in order to achieve high accuracy of the research. Results. The author conducted the analysis of the following datasets devoted to academic performance at higher and secondary educational institutions in a number of countries: Students' Performance in Portugal, E-learning Student Reactions and Students' Academic Performance. The purposes of the current study were to identify statistical correlations between social parameters of students and the level of their academic performance and to understand how academic performance is determined by the implementation of online learning and blended learning. The research findings suggest that mathematical statistics and data analysis methods allow to identify correlations between students' performance data and reveal hidden relationships which can be important for university staff. Conclusions. In conclusion, the author summarizes the results of evaluating the impact of the introduction of e-learning elements and some social parameters on students' academic performance. © 2020 Novosibirsk State Pedagogical University. All rights reserved.