TECHNOLOGY FOR PERSONALITIES SOCIALIZATION BY COMMON INTERESTS BASED ON MACHINE LEARNING METHODS AND SEO-TECHNOLOGIES

Abstract

Context. The socialization of individuals with common interests is caused by the need of most people to simplify some of the moments of life by reducing the time for their implementation. With the rapid growth of information, the human workload in society and the recent epidemics of the world, people are becoming isolated from the opportunity to communicate. And this is one of the important needs of human consciousness and self-realization. Therefore, there is an urgent need to be able to obtain a recommended list of similar people of common interest as a result of intelligent search of many relevant users of social networks through analysis of human faces in user photos (based on neural networks) and analysis of user information based on fuzzy search algorithms and Noisy model. Channel). Objective of the study is to develop technology for socialization of individuals based on SEO-technology and machine learning through the use of convolutional and Siamese neural networks to identify users and text analysis algorithms to select relevant users of future communication. Method. In the implementation of SEO-technologies selected fuzzy word search algorithms based on the Noisy Channel model algorithms for efficient distribution of textual information. During the implementation of machine learning, a convolutional neural network was developed to identify users of the system. Results. An intelligent system of socialization of individuals by common interests based on SEO-technology and machine learning methods has been developed. The work of two neural networks was implemented: convolutional and Siamese, which allowed to search for a human face in photos uploaded by the user and compare the found face with those already available in the database / Internet. This makes it possible to effectively identify the authenticity of the user and ensure that this user is not currently in the database, so it is potentially real. Using fuzzy search algorithms, Levenstein’s algorithm and the Noisy Channel model, an algorithm for analyzing and comparing user information was created, which for the current user forms a list of available users of the system, sorted by descending percentage of similarity and indicates how other users’ interests coincide. Conclusions. It was found that the algorithm implemented in the system for forming a sample of users is more efficient and accurate by about 25–30% compared to the usual Levenstein algorithm. Also, the implemented algorithm performs sampling approximately 10 times faster than the usual Levenstein algorithm.