Technique for classifying the social network profiles according to the psychological scales based on machine learning methods

Abstract

Abstract A technique for classifying the social network users and groups by psychological scales of the Ammon’s test has been developed. To analyze user profiles, we have used several types of artificial neural networks (support vector machine, linear regression, multilayer neural network and convolutional neural network), and for group analysis, we applied text classifiers (bag of words, weighted bag of words, continuous bag of words, skip-gram and fastText classifier). The scope of the technique is identifying deviations in the psychological state of users of social networks and monitoring these changes considering users’ groups to detect destructive influences. An experiment was carried out, as a result of which it was found that a multilayer neural network with an activation function of the ReLU type has the best accuracy.