A Spatial Flow Clustering Method Based on the Constraint of Origin-Destination Points’ Location

Abstract

With the development of mobile positioning technology, a large number of Origin-Destination (OD) flow data with spatial and temporal details have been produced. These OD flow could give us a great opportunity to research geographical phenomena such as spatial interaction and mobility patterns. The OD flow clustering approach is an effective way to explore the main mobility patterns of the objects. At the same time, similarity measurement plays a key role in OD flow clustering. However, most of the previous OD flow similarity measurement methods failed to make full use of the spatial information of the flow including spatial proximity and geometric similarity. In this paper, we considered both position information and geometric properties of OD flow and propose a new method to measure the spatial similarity between OD flows. Specifically, the proposed method sets the neighbor threshold with the length of OD flows and the parameter dynamically. Based on the constraint of the OD points' location, the directions of the flows are implicitly restricted. The sole-parameter has a practical value as it determines the maximum length difference and the maximum directional difference that can be tolerated between similar flows. The proposed method passed a simulation experiment with synthetic flows and a case study with 283,008 taxi trips in Beijing in one day. The results show that the proposed method can discover the dominant mobility pattern from a large number of flow data effectively. In the case study, the dominant flow clusters reveal the taxi mobility patterns of residents at different distances in Beijing. INDEX TERMS Mobility pattern, origin-destination flow, spatial similarity, spatial clustering.

Investigating Human Travel Patterns from an Activity Semantic Flow Perspective: A Case Study within the Fifth Ring Road in Beijing Using Taxi Trajectory Data

“…α is a size coefficient and the product of α and the shorter length equals the radius of the boundary circle. We select α = 0.3 which is consistent with the existing work [52,53]. The smaller SD ij is, the more similar the flows are.…”

“…The flow clustering process is shown in Algorithm 1. For more detailed parameter settings, please refer to [52].…”