Abstract:

The implementation of the state educational policy in the direction of identifying and supporting talents, career guidance and individualization of training forms a request for an analysis of factors affecting the educational achievements of students. However, existing analytics is limited to assessing the dependence of academic performance on personal (related to the student) and institutional (related to the educational institution) factors. The data source for such analytics is the data generated within the framework of the LMS systems of educational institutions, the data of psychological tests, as well as educational and socio-economic statistics. At the same time, an enormous amount of data is ignored, which allows us to record and evaluate the educational achievements of students - the data generated on the Internet and its key locus for modern youth - social networks. Currently existing attempts to use this data to individualize the learning process boil down to the use of LMS data from educational institutions, information about the learning process and progress to build a path within the course being studied. Meanwhile, the penetration of social media into the everyday and professional life of modern society actualizes the study of methods and tools for extracting and analyzing data about users of social networks - digital footprints. The study of the human digital footprint allows the analysis and modeling of its physiological, psychological and cognitive characteristics and the use of such a model for predicting, programming and managing the educational trajectory. In our case, it is important that the digital footprint of a person in social networks provides a fundamentally new opportunity for assessing and analyzing educational achievements of students - it allows you to evaluate their informal and informal educational paths and achievements that were previously invisible to the state and educational institutions. In addition, the structure and content of human activity in the Internet space is an important component of its digital footprint, which forms the need for a separate study of its impact on educational trajectories and achievements.

The purpose of this work is to identify the relationship between the level of academic performance of students and their subscriptions to thematic communities on the «VKontakte» social network. The study used data on the performance of 2864 students (1799 female, 1065 male). For each student, community subscription data is uploaded using the «VKonakte» API. We managed to find a pattern in the subscriptions of students with high and low academic performance. The paper presents thematic trends in the content consumed that are characteristic of successful and lagging students. Application of machine learning algorithms to a data set allows predicting the probability of student expulsion from a university with an accuracy of 0,63 - 0.86. This approach can be used for predictive learning analytics in order to retain students, when at the beginning of the semester, a teacher using the digital footprint of students from a social network, can make a prediction about the probability of getting low marks for the exam during the session.

Keywords:

Academic performance, social network, machine learning, digital footprint, predictive learning analytics.