期刊
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
卷 22, 期 7, 页码 4496-4506出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TITS.2020.3022612
关键词
Trajectory database; outlier detection; individual trajectory outliers; group trajectory outliers; taxi frauds; GPU computing
This paper introduces a new solution to identify trajectory outliers in taxi fraud detection, presenting a two phase-based algorithm to identify both individual and group outliers. The proposed GPU approach significantly boosts the runtime performance, demonstrating the effectiveness of the method through extensive experiments.
This paper addresses the taxi fraud problem and introduces a new solution to identify trajectory outliers. The approach as presented allows to identify both individual and group outliers and is based on a two phase-based algorithm. The first phase determines the individual trajectory outliers by computing the distance of each point in each trajectory, whereas the second identifies the group trajectory outliers by exploring the individual trajectory outliers using both feature selection and sliding windows strategies. A parallel version of the algorithm is also proposed using a sliding window-based GPU approach to boost the runtime performance. Extensive experiments have been carried out to thoroughly demonstrate the usefulness of our methodology on both synthetic and real trajectory databases. The results show that the GPU approach enables reaching a speed-up of 341 over the sequential algorithm on large synthetic databases. The efficiency of the proposed method to detect both individual and group trajectory outliers on a real-world taxi trajectory database is also demonstrated in comparison with baseline trajectory outlier and group detection algorithms. The results are very promising and show superiority of the proposed method both in reducing computational time and enhancing the quality of returned outliers. Finally, we prime our methodology and results for future refinement using deep learning methodologies.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据