Challenges and Opportunties in Taxi Fleet Anomaly Detection

To enhance fleet operation and management, taxi operators instrument their vehicles with GPS receivers and network connectivity to servers. Mobility traces from such large fleets provide significant information on commuter travel patterns, traffic congestion and road anomalies, and hence several researchers have mined such datasets to gain useful urban insights. The fleet companies, however, incur significant cost in deploying and maintaining their vast network of instrumented vehicles. Thus research problems, that are not only of interest to urban planners, but to the fleet companies themselves are important to identify, to attract and engage these companies for collaborative data analysis. In this paper, we show how GPS traces from a taxi company can be used to answer three different questions that are of great interest to the taxi operator. These questions are 1) What is the occupancy rate of the taxi fleet?, 2) Do taxi drivers often take inefficient routes when serving passengers, and 3) Are there a large number of taxi drivers who are traveling significantly faster than the posted speed limits? We provide answers to each of these questions using a 2 month dataset of taxi records collected from about 15,000 taxis located in Singapore. The goal of this paper is to stimulate interest in the questions listed above (as they are of high interest to fleet operators) while also soliciting suggestions for better techniques to solve the problems stated above.