Learning Individual Behavior Using Sensor Data: The Case of GPS Traces and Taxi Drivers
COM2 Level 4
Executive Classroom, COM2-04-02
(Joint work with Ramayya Krishnan and Yingjie Zhang)
Abstract:
The ubiquitous deployment of mobile and sensor technologies enables observation and recording of human behavior in physical (offline) settings in a manner similar to what has been possible to date in online settings. This provides researchers with a new lens through which to study and better understand previously unobservable individual decision-making processes. In the present study, using a Bayesian learning model with a rich data set consisting of approximately 2 million fine-grained GPS observations, we analyzed the decision-making behavior of 2,467 single-shift taxi drivers in a large Asian city with the objective of understanding key factors that drive the supply side of urban mobility markets. The data set includes detailed taxi GPS trajectories, taxi occupancy (i.e., whether the taxi is occupied or not) data, and taxi drivers' daily incomes. This capacity to use data where occupancy of the taxi is known is a distinctive feature of our data set and sets our work apart from prior work in the literature. The specific decisions we focused on pertain to actions drivers take to find new passengers after they have dropped off current passengers. In particular, we studied the role of information derivable from GPS trace data (e.g., where passengers were dropped off, where they were picked up, longitudinal taxicab travel history with fine-grained time stamps) observable by or made available to drivers in enabling them to learn the distribution of demand for their services over space and time. We found significant differences between new and experienced drivers in both learning behavior and driving decisions. Drivers benefit significantly from their ability to learn from not only information directly observable in the local market but also aggregate information on demand flows across markets. Interestingly, our policy simulations indicate that information that is noisy at the individual level becomes valuable when being aggregated across relevant spatial and temporal dimensions. Moreover, we found that the value of information does not increase monotonically with the scale and frequency of information sharing. Our results also provide important evidence that efficient information sharing can lead to a welfare increase among drivers due to potential market expansion. Efficient information sharing can bring, within the taxi market, additional income generating opportunities that could be unfulfilled. Overall, the present study not only explains driver decision-making behavior but also provides taxi companies with an implementable information-sharing strategy to improve overall market efficiency.
Biodata:
Prof. Beibei Li is an Associate Professor of Information Systems and Management at Carnegie Mellon University, Heinz College. She received her PhD in Information Systems from Leonard N. Stern School of Business at NYU. Prof. Li's research interests lie at the intersection between social and technical aspects of information technology. Recently, she has been focused on the emerging digitization of human behavior. In particular, the mobile, web and sensor technologies today allow researchers to examine not only "what people say", but also "what people do" in both online and offline environments. Prof. Li's recent research has been published in Marketing Science, Management Science, MIS Quarterly and ISR. And many top IS, Economics, Marketing and CS conferences. She is an associate editor of ISR. She is the recipient of NSF Award on Smart and Connected Communities, the Anna Loomis McCandless Chair Professorship at Carnegie Mellon University. In 2015, she received Junior Marketing Researcher Award, Google Faculty Research Award, Adobe Faculty Research Award, and Marketing Science Institute (MSI) Research Award.