METHODS AND TOOLS TO IDENTIFY PROMISING ENTRANTS IN SOCIAL NETWORKS

Tomsk state University created the Consortium of universities for joint research in the field of big data Analytics. The consortium launched the project “University of open data” (http://data.tsu.ru), which brought together specialists in the field of computer technology and other Sciences - sociologists, political scientists, philologists, psychologists, biologists, geneticists. Tools available for uploading with the API custom data from “Vkontakte” for user of the project . Also we used the program “1C:School Psychodiagnostics” as a tool for collecting, cleaning, structuring and storing data in the study. The first area of research was related to testing of the hypothesis on the relationship between the interests of the entrants in social networks and choice of training programmes. We developed a model predicting choice of the direction of preparation of the applicant was tested on a large sample of the users “Vkontakte” (126 000). The accuracy of prediction equaled to 0.82 for the humanities, of 0.76 for the mathematical, of 0.69 for the natural sciences. The second direction of studies is connected with the automatic validation of methods of analysis and text classification to determine the scientific and professional interests of the applicant. The results of the study confirmed the possibility of the classification of texts in the HUMANITIES, natural and mathematical science with precision of 0.64. The third direction of research is devoted to studying the potential of social networks to identify key talent to diagnose psychological qualities of entrants through analysis of their subscriptions in a social network. To solve the problem of binary classification used support vector machines. The accuracy of a predictive model based on this method were as follows: 0.7 for identification of children with high intelligence; 0.7 for creative; 0.72 for children with high personal motivation. In the fourth direction was analyzed the social graph (SNA) for entrants with high and low test results to determine the level of intellectual development. The result failed to reveal significant differences in the number of friends pupils with low and high level of intelligence. Visualization of the network of intellectuals has shown the existence of a single network virtual connections, despite the fact that it includes entrants from 54 educational institutions of several cities in Tomsk region. Also the analysis of social networks of intellectuals helped to identify opinion leaders - entrants with the greatest number of people in your network that will improve the effectiveness of communication of the University with the task applicant. At the current stage of research it is possible to design a predictive model to identify promising entrants through the projection of the target model of the graduate of TSU. The composition of the model should include the following criteria: · a high level of intelligence and interest to science as the projection quality of a «developed personality» of the target model of a graduate of TSU; · high level of creativity as the projection «create new technological and social reality.» Assessment of entrant’s potential according to two criteria of this model will allow universities to identify high school entrants with a full or partial set of desired qualities, and to organize their early engagement in the University environment for the formation of interest to the educational program of the University and prepare for entry and training.

Keywords

Authors

References