Abstract
Abstract. Taxi trajectory data contains the detailed spatial and temporal traveling information of urban residents. By using a clustering algorithm, the hotspots’ distributions of pick-up and drop-off points can be extracted to explore the patterns of taxi traveling behaviors and its relationship with urban environment. Comparing with traditional methods that determine hotspots at a relatively large scale, we propose an approach to detect small-scale hotspots, so called docking points, to represent the local clusters in both sparse and dense stops areas. In this method, we divide the research area into grids and extract the docking points by finding local maximums of a certain range. The extracted docking points are classified into five levels for the subsequent analysis. Finally, to uncover detail characteristics of taxi mobility patterns, we analyze the distributions of docking points from three aspects – the overall, by day of the week, and by time of the day.
Highlights
As an essential part of urban public transportation, taxis provide point-to-point services to urban residents
According to the concept of ‘docking points’ in Section 4.2, this paper proposes local maximum density (LMD) method based on grids to extract the docking points
The main difference between DBSCAN and LMD method is that DBSCAN detects clusters by connecting core points and corresponding EPS areas, while our method aims at extracting a small E range area around a local maximum grid
Summary
As an essential part of urban public transportation, taxis provide point-to-point services to urban residents. Taxi GPS trajectories data records abundant and detailed spatial-temporal traveling information of passengers, which reflects the traveling behaviors of urban residents to a certain degree (Wang et al, 2015; Zhang et al, 2015). Considering that taxis are point-to-point services, small-scale hotspots are in existence. A small-scale hotspots detection method is needed to uncover abundant detailed patterns of the mobility behaviors of people. This paper proposes a novel density-based approach to detect small-scale local hotspots (defined as ‘docking points’) for pickup and drop-off points of taxis. The grids in neighborhood E are classified into corresponding local maximums. Based on these procedures of the proposed clustering method, it is named as local maximum density (LMD) approach.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
