以社群網站照片主色相預測人格特質之研究

Predicting Personality Traits Through Dominant Colors of Images on Social Media

陳俞君(Yu-Chun Chen) ； 許秉瑜(Ping-Yu Hsu) ； 莊雅筑(Ya-Chu Chuang)

《電子商務學報》 預刊文章 Pp. 1-24

https://doi.org/10.6188/JEB.202307/PP.0004

摘要

人格特質被廣泛應用在各領域，透過人格特質了解不同群體利於達到個人化。在眾多顏色人格特質測試中，Lüscher Color Test（LCT）獲得最廣泛使用，但尚未有研究將Lüscher顏色與五大人格特質關聯。根據繪畫與顏色心理學，本研究嘗試萃取社群網站使用者照片之主色相（Dominant color），將其與Lüscher顏色連結，建立機器學習模型預測使用者之五大人格特質。實驗結果以Random Forest演算法所建構的模型，單一人格特質最高準確率可達95％，五項人格特質平均為79％。本研究之貢獻在於提出kmeans++動態萃取社群網站使用者風景照之主色相，並驗證主色相與Lüscher顏色連結可以準確預測使用者之五大人格特質。

關鍵字

Lüscher顏色心理學 ； 五大人格特質 ； 主色調 ； 社群網站照片 ； 機器學習

並列摘要

Personality traits are widely used. Understanding different groups through personality traits is beneficial to achieving personalization. Among the many color personality tests, the Lüscher Color Test (LCT) is the most widely used. However, no research has linked Lüscher's color with the Big Five personality traits. According to the psychology of painting and color, the present study attempts to extract the dominant color of social media site users' images, connect it with the Lüscher's color, and establish a machine learning model to predict the big five personality traits of users. The experimental results show that the model constructed by the Random Forest algorithm is the most promising. The highest accuracy rate of a single personality trait can achieve 95%, and the average of five personality traits is 79%. The contribution of the present study is to propose that kmeans++ dynamically extracts the dominant color of landscape images of social media site users, and to verify that the connection between dominant color and Lüscher's color can accurately predict the Big Five personality traits of users.

並列關鍵字

Lüscher color psychology ； big five personality traits ； dominant color ； social media images ； machine learning