Who are happier? Spatio-temporal Analysis of Worldwide Human Emotion Based on Geo-Crowdsourcing Faces

Geotagged social media data provides unprecedented opportunities and meaningful aspects of human analysis in the era of volunteered geographic information (VGI). Previous studies have examined users‘ emotions shared on these media, while most of them focused on text-based data and ignored diverse images. In this paper, we used a huge global scale image dataset: YFCC100, to extract emotions from photos and to describe the worldwide geographic patterns of human happiness. Two indices of Average Smiling Index (ASI) and Happiness Index (HI) are defined from different perspectives to describe the degree of human happiness in a specific region. We computed the spatio-temporal characteristics of facial expression-based happiness on a global scale and linked them to some demographic variables (ethnicity, gender, age, and nationality). After that, the robust analysis was made to ensure our results are reliable. Results are in accordance with some previous studies and common sense, for example, White and Black are better at expressing happiness than Asian, women are more expressive than men, and happiness expressed varies across space and time. Our research provides a novel methodology for emotion measurement and it could be utilized for assessing a region‘s emotion conditions based on geo-crowdsourcing data. Robust analysis results on our dataset indicate that our approaches are reliable and could be implemented in research of human emotions.