Abstract:
The academic performance of students at Bauman Moscow State Technical University and its correlation to their activity in the VKontakte social network are considered. Machine learning methods are used to identify distinct performance paths reflecting the dynamics of educational achievement. The subscription lists of students are analyzed to identify marker communities characterizing the predominance of specific performance categories. Graph-theoretic clustering is applied to reveal structural groups of student interests. For each path, the stochastic vectors of interest shares across the community clusters are constructed and then used to identify the clusters having a statistically significant relation to particular performance paths. The results indicate a correlation between the digital behavior and academic outcomes of students, which contributes to the development of performance prediction models considering student interests in a social network.
Fulltext:
PDF file (2446 kB)