The emergence of social media provides a new source of sensing society. In this thesis, a novel event detection method for detecting real-world events based on population sentiment orientation (PSO) from social media check-in data is proposed. The method is mainly composed of sentiment analysis, spatial-temporal analysis, and event extraction. The hypothesis guiding this research is that social events change PSO in the dimension of time and space. The ratio of the number of positive and negative records is chosen to indicate the PSO within a specified period and geographical area. First, different sentiment classification methods are compared and a sentiment classification model is trained. Second, spatial-temporal analysis empowers the method to detect multi-scale events in terms of time and space dimension. Specifically, the method can detect events from nationwide festivals to local activities and from annual-scale to day-scale. Furthermore, time series analysis and spatial clustering can interpret the social phenomenon of the event by mining spatial-temporal patterns. At last, A Word Cloud is used to visualize the extracted high-frequency event keywords for visually identifying the event. For testing the method, a case study is conducted using Sina Weibo data in Shanghai, China, 2014. Events such as Chinese New Year, Mid-Autumn Festival and concerts have been successfully detected. Besides, the interpretation ability of this method is also tested. Main reasons for negative microblogs on the New Year’s Eve are successfully extracted. Traffic as one of the biggest reasons from the result, the spatial pattern of it has been discovered. Related microblogs are mainly distributed at sites of public transportation like Shanghai Pudong international airport or railway station.     «

The emergence of social media provides a new source of sensing society. In this thesis, a novel event detection method for detecting real-world events based on population sentiment orientation (PSO) from social media check-in data is proposed. The method is mainly composed of sentiment analysis, spatial-temporal analysis, and event extraction. The hypothesis guiding this research is that social events change PSO in the dimension of time and space. The ratio of the number of positive and negativ...     »