Bike-sharing source-sink space recognition based on riding flow density clustering method

Objectives: The shared bike is an important tool for residents to connect to public transport and travel a short distance, but its uneven space-time distribution has caused inconvenience to operators and users, especially in the morning and evening rush hours of commuting on weekdays. Therefore, it is necessary to identify the spatial divergence area (source) and convergence area (sink) of shared bikes, and carry out targeted management and scheduling of their spatial source-sink area. However, most studies aggregate flow data into polygon units, which lose the precision of flow data and cannot provide accurate spatial location of source and sink areas. This paper uses a method combining spatial thinning with flow density clustering to effectively identify the spatial location of the source and sink area of bike-sharing flow. Methods: Spatial community detection algorithm and density-based clustering method for OD riding flow were used to identify the typical source and sink regions of shared bikes. The first thing is to clean the original dataset, including deleting duplicate orders, orders with missing attributes and abnormal orders, and performing coordinate matching. After that, using the fast Fourier transform algorithm to test the time periodicity of the dataset. Then the fast unfolding algorithm is used to detect the spatial communities and dilute the order data of shared bikes on this basis. Finally, define the spatial distance of bike-sharing flow based on the road network, and then use the shared K nearest neighbor flow clustering method to identify typical source and sink regions. Results: (1) The curve of bikesharing timely orders on weekdays has obvious M-shaped characteristics, while the curve of weekends is more similar to the ladder type. At the scale of traffic analysis zone (TAZ), the flow of shared bikes is mainly adjacent flow and internal flow; (2) The boundary of bike-sharing community is significantly more constrained by natural barriers than administrative divisions, and the highvalue bike-sharing usage areas are more concentrated in the morning peak than evening; (3) During the morning rush hours of commuting on weekdays, the unidirectional flow clusters are mainly "residence to employment", and the convergent and divergent areas show the spatial distribution characteristics of "one area, one line and many points"; In the evening peak hours, the flow clusters are mainly the pattern of subway station to residential area, and the intermodal characteristics of "shared bikes-subway station-shared bikes" are prominent. Conclusions: Based on the perspective of flow space, this paper analyzes the characteristics of residents' short-distance shared travel. The proposed method effectively identifies the spatial source and sink area of shared bikes, which fills the smallest link in the study of urban internal mobility links, and provides support for the optimal allocation of shared bicycle space.