F-Trail: Finding Patterns in Taxi Trajectories

Publication Type

Conference Proceeding Article

Publication Date

4-2013

Abstract

Given a large number of taxi trajectories, we would like to find interesting and unexpected patterns from the data. How can we summarize the major trends, and how can we spot anomalies? The analysis of trajectories has been an issue of considerable interest with many applications such as tracking trails of migrating animals and predicting the path of hurricanes. Several recent works propose methods on clustering and indexing trajectories data. However, these approaches are not especially well suited to pattern discovery with respect to the dynamics of social and economic behavior. To further analyze a huge collection of taxi trajectories, we develop a novel method, called F-Trail, which allows us to find meaningful patterns and anomalies. Our approach has the following advantages: (a) it is fast, and scales linearly on the input size, (b) it is effective, leading to novel discoveries, and surprising outliers. We demonstrate the effectiveness of our approach, by performing experiments on real taxi trajectories. In fact, F-Trail does produce concise, informative and interesting patterns.

Discipline

Software Engineering | Transportation

Research Areas

Software Systems

Publication

Advances in Knowledge Discovery and Data Mining: 17th Pacific-Asia Conference on Knowledge Discovery and Data Mining (PAKDD 2013)

First Page

86

Last Page

98

ISBN

9783642374531

Identifier

10.1007/978-3-642-37453-1_8

Publisher

Springer Verlag

City or Country

Gold Coast, Australia

Additional URL

http://dx.doi.org/10.1007/978-3-642-37453-1_8

Link to Full Text

Share

COinS