Thursday, February 25, 2010
12:00 pm - 1:00 pm
PLACE: Gates Center 8102
Rajesh Balan, School of Information Systems,
Singapore Management University
Real-Time Information Services for a Large Taxi Fleet
In this talk, I will describe two information services designed for a real-world fleet of more than 15,000 taxicabs. The two services are (1) a trip information service that provides passengers with the expected fare and duration of the taxi ride they are planning to take, and (2) a demand prediction service that advises drivers where to go to increase their chances of finding a passenger. Both services were designed to operate in real time. I will describe both the algorithms developed to support these services as well as the evaluation of our developed services. Our results, tested on a dataset of over 2.6 billion records (from 7 months of data collected from the taxi provider), demonstrate that both services are (a) quite accurate, and (b) capable of real-time performance. In addition, I will also provide insights in choosing the right amount of history in making predictions, the usefulness of online learning techniques, and the use of heuristic thresholds to improve the quality of results. Finally, if time permits, I will also discuss the challenges of working with large amounts of spatial data; namely, dealing with noisy data, spatial databases, and large amounts of data.
Rajesh Balan obtained his Ph.D. from Carnegie Mellon University in 2006 and joined the School of Information Systems at the Singapore Management University in 2006 immediately afterwards. He is thrilled to be returning "home" to speak at the SDI seminar. His research interests are quite eclectic and include developing novel mobile and pervasive applications, investigating and building massively multiplayer game infrastructures, and understanding and analyzing the issues related to distributed software engineering.
Visitor Host: M. Satyanarayanan
Visitor Coordinator: Tracy Farbacher, firstname.lastname@example.org
SDI / LCS Seminar Questions?
Karen Lindenfelser, 86716, or visit www.pdl.cmu.edu/SDI/