Course project
Abstract: GPS-enabled devices have become an integral part of everyday life, leading to a massive amount of mobility data being generated. In addition to being high in volume, mobility data is high-dimensional. This makes it challenging to discern underlying mobility patterns in a raw spatiotemporal form. Therefore, a representation that clearly accounts for spatial and temporal mobility patterns is needed. This project focuses on applying the graph representation technique for a sparse representation and analysis of mobility data. We analyze the mobility traffic traces from GPS traces of 182 users in Beijing. First, we show that mobility data can be represented using a directed graph. The resulting graph has a degree distribution exhibiting the power-law distribution of human mobility found in the literature. Second, we extract interesting locations based on the graph’s properties. Finally, we study the potential application of GSP for analyzing the temporal dynamics of mobility by considering an average movement of users over some small time window in a given node as a signal to the graph.