Modeling of Personality Traits based on Demographic Data on Multi-Races Samples of Ages from 13 to 50 Years Old: Investigating the Effect of Race on Model

The current method to predict personality in personality-based recommender systems is by using Personality Extraction from Text (PET). Since this method has a flexibility weakness, a new method that is based on demographic data is proposed. The objective of this paper is to study the effect of race on the resulted model. In this study, we compare models obtained from International data, which comprise many races, and SE Asian data containing only one race. The results of the study reveal that races do influence the accuracy of the model. The International models are less accurate than those of SE Asian models are. We suspect that this happens because each race has its own personality level. This claim is supported by previous studies on personality differences across nations. These studies have found that personality differences across nations do exist. Therefore, we hypothesize that the more homogenous the data in terms of race, the more accurate the model.