Exploring the Urban Region-of-Interest Through the Analysis of Online Map Search Queries
Ying Sun,Hengshu Zhu,Fuzhen Zhuang,Jingjing Gu,Qing He
DOI: https://doi.org/10.1145/3219819.3220009
2018-01-01
Abstract:Urban Region-of-Interest (ROI) refers to the integrated urban areas with specific functionalities that attract people's attentions and activities, such as the recreational business districts, transportation hubs, and city landmarks. Indeed, at the macro level, ROI is one of the representatives for agglomeration economies, and plays an important role in urban business planning. At the micro level, ROI provides a useful venue for understanding the urban lives, demands and mobilities of people. However, due to the vague and diversified nature of ROI, it still lacks of quantitative ways to investigate ROIs in a holistic manner. To this end, in this paper we propose a systematic study on ROI analysis through mining the large-scale online map query logs, which provides a new data-driven research paradigm for ROI detection and profiling. Specifically, we first divide the urban area into small region grids, and calculate their PageRank value as visiting popularity based on the transition information extracted from map queries. Then, we propose a density-based clustering method for merging neighboring region grids with high popularity into integrated ROIs. After that, to further explore the profiles of different ROIs, we develop a spatial-temporal latent factor model URPTM (Urban Roi Profiling Topic Model) to identify the latent travel patterns and Point-of-Interest (POI) demands of ROI visitors. Finally, we implement extensive experiments to empirically evaluate our approaches based on the large-scale real-world data collected from Beijing. Indeed, by visualizing the results obtained from URPTM, we can successfully obtain many meaningful travel patterns and interesting discoveries on urban lives.