Path inference is an essential component for many location based services. In this paper, we study the problem of inferring vehicle moving paths from noisy and incomplete data captured by GPS devices mounted on vehicles. We propose a collaborative filter model to incorporate both static and dynamic context information to achieve highly accurate path inference. A tensor decomposition technique is adopted to extract context-aware spatial and temporal features from the location data with minimal a prior information about the underlying roads such as the path lengths. We evaluated our framework using a large scale real world dataset, which has one-month location data from more than eight thousand taxis in Beijing. The evaluation results show that our method outperforms state-of-the-art techniques.
CITATION STYLE
Wang, H., Wen, H., Yi, F., Li, Z., & Sun, L. (2016). Tensor filter: Collaborative path inference from GPS snippets of vehicles. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9798 LNCS, pp. 103–115). Springer Verlag. https://doi.org/10.1007/978-3-319-42836-9_10
Mendeley helps you to discover research relevant for your work.