Fraud detection from Taxis' driving behaviors

Siyuan Liu, Lionel M. Ni, Ramayya Krishnan

Research output: Contribution to journalArticlepeer-review

65 Scopus citations


Taxi is a major transportation in the urban area, offering great benefits and convenience to our daily life. However, one of the major business fraud in taxis is the charging fraud, specifically overcharging for the actual distance. In practice, it is hard for us to always monitor taxis and detect such fraud. Due to the Global Positioning System (GPS) embedded in taxis, we can collect the GPS reports from the taxis' locations, and thus, it is possible for us to retrieve their traces. Intuitively, we can utilize such information to construct taxis' trajectories, compute the actual service distance on the city map, and detect fraudulent behaviors. However, in practice, due to the extremely limited reports, notable location errors, complex city map, and road networks, our task to detect taxi fraud faces significant challenges, and the previous methods cannot work well. In this paper, we have a critical and interesting observation that fraudulent taxis always play a secret trick, i.e., modifying the taximeter to a smaller scale. As a result, it not only makes the service distance larger but also makes the reported taxi speed larger. Fortunately, the speed information collected from the GPS reports is accurate. Hence, we utilize the speed information to design a system, which is called the Speed-based Fraud Detection System (SFDS), to model taxi behaviors and detect taxi fraud. Our method is robust to the location errors and independent of the map information and road networks. At the same time, the experiments on real-life data sets confirm that our method has better accuracy, scalability, and more efficient computation, compared with the previous related methods. Finally, interesting findings of our work and discussions on potential issues are provided in this paper for future city transportation and human behavior research.

Original languageEnglish (US)
Article number6557504
Pages (from-to)464-472
Number of pages9
JournalIEEE Transactions on Vehicular Technology
Issue number1
StatePublished - Jan 2014

All Science Journal Classification (ASJC) codes

  • Automotive Engineering
  • Aerospace Engineering
  • Electrical and Electronic Engineering
  • Applied Mathematics


Dive into the research topics of 'Fraud detection from Taxis' driving behaviors'. Together they form a unique fingerprint.

Cite this