Journal of Psychiatry

Predicting the scores (personality type) of Myers-Briggs type indicator based on the scores of Holland inventory using MLP neural network: A new approach to personality assessment

27^th International Conference on PSYCHIATRY & PSYCHOLOGY HEALTH

June 18-19, 2018 Paris, France

Ali Ansari

Islamic Azad University Evaz, Iran

Posters & Accepted Abstracts: J Psychiatry

Abstract:

Each individual has a combination of personality traits which may include extraversion/introversion, etc. and if we measure these traits with numeric values, we�??ll be able to perform mathematical computations on them. Among the most widespread computations are descriptive statistics, correlation and also regression analysis. In this research, we suggested a new method for predicting a set of traits, based on another set. We used a dataset of 300 samples in which each sample includes the scores of Holland inventory and the scores of MBTI Type Indicator. After statistical analysis, results showed that the correlations are not statistically significant. Then, we examined another tool called �??MLP (Multilayer Perceptron) neural network�?�. The dataset has been used to train and test the MLP NN and Root-mean-square error (RMSE) and precision score of the trained network were calculated. Regarding the correlation coefficients obtained in the previous step and also the scatter plots, there is no linear relationship between the scores of these two scales. However, the MLP NN which has been trained using hyperbolic tan function as activation function, had higher predictive power and can be used to predict output measures (of MBTI Type Indicator) based on input measures (of Holland inventory). Based on the results, in order to explain relationships between mental attributes and to predict one attribute based on the others, NNs (in particular, MLP) are more powerful than traditional methods, when (1) the relationships are non-linear and (2) we need numeric outputs and not subjective interpretations based on correlations.