User Group Classification Methods Based on Statistical Models

The fundamental difficulty of building an information model of a target object in social networks is that a large number of characteristics (several tens) are used in the description of objects in social networks, described by all conceivable types of data: numbers, score estimates of qualitative characteristics, texts, symbols, video and audio information. Obviously, such non-additive data types cannot be used to construct any integral criterion for the specification of the target object. To solve this problem, the article introduces the concept of “vector of target search”. The general idea for solving this problem, proposed by the authors, is to convert physical characteristics into relative dimensionless quantities with normalized values from 0 to 1. The authors have implemented modern promising ideas in the development of intelligent information technologies, such as: computer training of intelligent agents using illustrative examples from the training sample, agent-based technologies for working with Big Data, the method of wave scanning of social networks when searching for target objects. The implementation of wave scanning of social networks during agent search of targets significantly reduces the computing power required to implement the search process and reduces the amount of “noise” in agent collections. Authors developed a method of marking a single object of a social network to solve the problems of streaming classification of objects in the interests of various groups of researchers, including solving the problems of targeted attraction of applicants to a higher educational institution.