Urban activity summarization with geo-tagged social media data

Information mining from large databases has become an important research topic for researchers in computer science and information technology. Data mining has been attractive to many researchers in different fields. Among these fields, urban activity summarization aims at modeling people's typical activities at different locations and time in a city. With the ever-increasing urbanization process, urban activity summarization is widely recognized as a crucial socioeconomic task. Previously, it was difficult to be done due to the lack of real-life geo-tagged social media (GTSM) data. In recent years, with the development of social media, such as tweeter and Weibo (widely used in China and similar with tweeter), there are sufficient acceptable data for solving this task. There are some progress made on the studies of geographical topics based on GTSM data, but their high computational costs and strong distributional assumptions prevent the release of GTSM data energy. In order to solve this problem, we propose a model of urban activity summarization. This model is based on the method of kernel density estimation to find out the spatiotemporal hot spots of people's activities and maximize the scarcity of GTSM data. Not only that, we have greatly reduced the time complexity by subdividing the space. Finally, we evaluate the validity of the model using the Weibo data, by comparing the results of the model prediction and the actual results, and give an answer to the proposed activities under a given time and place, as well as the recommendation of a given activity.