MBTI Personality Type Prediction Model Using WZT Analysis Based on the CNN Ensemble and GAN

Abstract

The Myers-Briggs Type Indicator (MBTI) personality analysis is a well-known method used to perform personality diagnosis of adolescents after a natural language processing analysis using social media data. However, directly using image data to analyze a personality type is insufficient. This research uses directly drawing image data presented by the adolescent using the Wartegg-Zeichen Test (WZT) and predicts MBTI personality types by mixing various representations of the convolutional neural network (CNN) and generative adversarial network (GAN) models. The study aims to analyze the data from 808 junior high school students. We employ a binary approach that divides MBTI personality types into four classes and presents improved prediction performance using the CNN ensemble and GAN techniques. As a result, the initial average predicted value of the CNN is 17.2%, but the average predicted value in the method using the final GAN is 27.2%, indicating an increase of 10%. This study is the first to predict various personality types and automatically analyze drawing images, expressed in the WZT of adolescents based on the deep learning model. Recent deep learning technologies evolve daily, and this study aims to create many opportunities for deep learning applications.