Mining Emotional Geography Features Based on Chinese Weibo Data

Emotion is a kind of geography knowledge existing in space and time, but its acquisition is difficult. Location social networks provide high-quality data sources for emotion measurement. Based on approximately 3.45 million Sina Weibo data with geo-locations, this paper tweeted between September 19-25, 2016 and October 1-7, 2016. Firstly, through text cleaning, Chinese word segmentation and sentiment analysis based on dictionaries and grammar rules, the sentiment tendency of each Weibo text was scored. Then, the emotion values were aggregated at city level and spatial distribution characteristics of city Weibo emotions were detected. Finally, hot-spot and cold-spot spatiotemporal patterns and trends of Weibo emotions were analyzed at hexagon grid level. The research shows that the overall Weibo emotion tendency in China is positive and spatial distribution of individual emotions is random. At city level, the spatial distributions of emotions is extremely uneven and presents sub-regional distribution pattern, which high-high or low-low clusters co-occur with low-high or high-low outliers in local areas and these local differentiations are more significant in September. At hexagon grid level, emotion patterns are opposite between the eastern and western regions of China. The eastern regions mainly present emotional hot-spot patterns, and September and October have different types of hot-spot patterns. In the western regions, emotional cold-spot patterns are prominent, and compared with September, cold-spot patterns of the National Day holiday tend to strengthen. The research reveals the diversities and differences of space-time distribution of Weibo emotions, which has instructive significance for analyzing the happiness of Chinese residents and social and economic development planning.