Trajectory Outlier Detection Based on Multi-Factors

Lei ZHANG  Zimu HU  Guang YANG  

IEICE TRANSACTIONS on Information and Systems   Vol.E97-D   No.8   pp.2170-2173
Publication Date: 2014/08/01
Online ISSN: 1745-1361
DOI: 10.1587/transinf.E97.D.2170
Type of Manuscript: LETTER
Category: Data Engineering, Web Information Systems
trajectory,  outlier detection,  multi-factors,  canonical correlation analysis,  

Full Text: PDF>>
Buy this Article

Most existing outlier detection algorithms only utilized location of trajectory points and neglected some important factors such as speed, acceleration, and corner. To address this problem, we present a Trajectory Outlier Detection algorithm based on Multi-Factors (TODMF). TODMF is improved in terms of distance-based outlier detection algorithms. It combines multi-factors into outlier detection to find more meaningful trajectory outliers. We resort to Canonical Correlation Analysis (CCA) to optimize the number of factors when determining what factors will be considered. Finally, the experiments with real trajectory data sets show that TODMF performs efficiently and effectively when applied to the problem of trajectory outlier detection.