Application of machine learning methods for preventive identification of deviant groups of adolescents

Abstract

The article is devoted to the problem of identifying deviant groups studying in educational institutions. The article substantiates the relevance of this problem. Recent incidents and possible reasons for their occurrence are analysed. The system for preventive detection of deviant groups of teenagers is offered. Categories of deviant groups of adolescents are identified: potential shooters; potential bullies; potential students capable of laying hands on themselves. It is suggested to measure anxiety and aggressiveness in adolescents in order to detect deviation of adolescents. The ways of measuring anxiety and aggressiveness are considered. It is suggested to use profiling as an objective method of assessment. Profiling by school psychologists is rather difficult to implement due to their limited number in educational institutions. To automate the process and exclude the subjective factor, it is suggested to use online profiling. Within the framework of online profiling, the objects of research are the pages of social networks used by students. Criteria for determining anxiety and aggressiveness based on information from social network pages are proposed. A toolkit for partial automation of the process of data collection from the pages of social networks of students of educational institutions is proposed. The use of the method of clustering k-means clustering method to identify deviant groups of students. An example of detection of teenagers with deviations using specially generated synthetic data is given.

Keywords

• school-shooting, columbine, aggressiveness assessment, anxiety assessment, online profiling, profiling of vkontakte social network pages, unsupervised machine learning, cluster analysis, k-means method.