| Abstrak/Abstract |
Several researchers have used posts on social media to estimate the personality traits of the authors. A personality- based recommender system that applies the method to predict the user's personality traits requires the user to write a status of a certain length. The problem with this is that not everyone writes a status on their social media accounts. Therefore, such a recommender system cannot be used for everyone. To solve this problem, we propose a new method of indirectly predicting personality traits which is based on demographic data. To be able to do this, a personality model was needed that relates demographic data to personality traits. As many as 325 personality models were created, of which there were 65 models for each of the following traits namely, agreeableness, emotional stability, intellect, extraversion, conscientiousness. We used three criteria to select the model to be used, namely demographic data that does not change in the course of one’s life, it does not have too many categories and the model has quite good accuracy. Based on the above criteria, we chose a model consisting of a combination of age group and gender. Another reason for choosing this model was the findings of previous researchers which state that there is a very close relationship between ages - gender and personality traits. The personality model reveals that each age group – gender cohort has specific personality except for adulthood female and middle age female who have the same personality. To be able to recommend certain items to users of the recommender system, one more model was needed, namely a preference model that connects personality traits with preferences for fashion styles. The preference model shows that only male with low emotional stability and high intellects likes Natural and Masculine fashion style. Meanwhile, male and female with other personalities like Elegant Chic and Natural fashion style. |
