Exploring the Potential of Combining Taxi GPS and Flickr Data for Discovering Functional Regions

The increasing deployment of GPS-enabled devices is leading to an increasing availability of trace data with various applications especially in the urban environment. GPS-equipped taxis have become one of the main approaches of collecting such data. However, we have realized two problems that may limit the effectiveness of use of these taxi GPS data in some applications. Firstly, taxi trajectories represent a very small portion of urban mobility in most of the cities. As a result, without other considerations important information that could come from non-taxi users is excluded. Secondly, advanced applications are built on the analysis of these traces and the context of the movement which is generally obtained from a set of points of interest (POIs). However, considering that POIs are predetermined, we argue that they are a very limited representation of the movement context. The approach we suggest supplements taxi trajectories with crowd-sourced data in an application to discover different functional regions in a city. We cluster the taxi pick-up and drop-off points to discover regions, then semantically enrich these regions using data from Flickr photos and determine the functions of the regions using this semantic information. The evaluation of the approach we performed using large datasets of taxi trajectories and Flickr photos allowed us to identify the potential and limits of the approach which are discussed in this paper.