Recognizing Social Function of Urban Regions by Using Data of Public Bicycle Systems

Obtaining the classification of urban functions is an integral part of urban planning. Currently, public bicycle systems are booming in these years. It conveys human mobility and activity information, which can be closely related to the social function of an urban region. This paper discusses the potential use of public bicycle systems for recognizing the social function of urban regions by using one year's rent/return data of public bicycles. We found that rent/return dynamics, extracted from public bicycle systems, exhibited clear patterns corresponding to the urban function classes of these regions. With seven features designed to characterize the rent/return pattern, our method based on Smooth support vector machine (SSVM) is proposed to recognize social function classes of urban regions. We evaluate our method based on the large-scale real-world dataset collected from the public bicycle system of Hangzhou. The results show that our method can efficiently recognize different types of urban function areas. Classification results using the proposed SSVM achieved the best classification accuracy of 96.15%.