Predicting the results of the 16-factor R. Cattell test based on the analysis of text posts of social network users

We investigated the possibility of automating the prediction of the 16-factor personality traits by R. Cattell from text posts of social media users. The proposed new method of automating the evaluation of R. Kettell's 16-factor personality test traits includes language models and neural networks. Implementation of the method involves several steps. At the first step text posts are extracted from user accounts of social media, pre-processed with language model RuBERT and previously trained over a full-connected neural network. The result of this step is a normalized empirical distribution of the posts by the previously introduced classes for each user. Subsequently, based on the distribution of user posts the evaluation of the expression of psychological features of the user is made with the help of support vector machine, random forest and Naive Bayesian classifier. The final data set for model building and further testing their performance was made up of 183 respondents who took the R. Cattell test, with links to their public social media accounts. Classifiers predicting results for six factors (A, B, F, I, N, Q1) of R. Cattells 16-factor personality test were constructed. The results can be used to create a prototype of automated system for predicting the severity of psychological features of social media users. Results of work are useful in the applied and research systems connected with marketing, psychology and sociology, and also in the field of protection of users from social engineering attacks.