DETERMINING OF PERSON’S CHARACTER TYPE BY CONVOLUTIONAL NEURAL NETWORK (USING THE MMPI METHODOLOGY)

Abstract

Background. This paper raises the problem of diagnosing character types (CT) using the MMPI test. The usual MMPI test consists of 600 questions and takes at least 1.5 hours, where the employer depends on the self-esteem of the test taker, and the retest reliability of the MMPI has not been established. Materials and methods. This paper researches the possibility of determining one of the eight CT by MMPI by a photo with trained convolutional neural network (CNN). To do this, using theoretical and methodological analysis, the most accurate tool for measuring the typology of personality traits was selected, after which a dataset was collected with photos of people who agreed to participate in the experiment and passed double testing. Next, a neural network was trained on the resulting dataset consisting of 40,581 photos. Results. The convolutional neural network (CNN) created by us correctly recognized 30 % of the 8116 photos provided to it with higher speed. Conclusions. Obtained results are comparable with estimated accuracy of paper testing and show the prospects of this approach.

Keywords

• personality traits
• testing
• psychodiagnostics
• normality and pathology of personality
• neural network