TY - JOUR

T1 - Modeling taxi driver search behavior under uncertainty

AU - Zheng, Z.

AU - Rasouli, Soora

AU - Timmermans, Harry J.P.

PY - 2021/1

Y1 - 2021/1

N2 - This paper develops a behavioral model of taxi driver passenger search behavior under uncertain conditions. We assume that, with a particular decision horizon in mind, a taxi driver chooses a particular strategy to search passengers. We differentiate between random search, maximum anticipated pick-up probability search, and maximum anticipated revenue search strategies. The probability of choosing a particular strategy is proportional to the weighted sum of the rewards associated with the possible outcomes of the decision, where the weights represent the driver’s beliefs, i.e. subjective probabilities, about the outcomes of the decision. Probability weightings are used to reweigh the belief set to capture drivers’ optimistic vs. pessimistic attitudes with respect to the uncertain outcomes. The reward function consists of two components: the anticipated monetary rewards of a decision within a certain time horizon, and the anticipated information value which captures the reduction in uncertainty as taxi drivers learn about their environment. The parameters of the value/reward function and the probability weighting are estimated using observed taxi trajectories derived from 1.5 million taxi global positioning system records. Results show that the probability weighting function signals an overall pessimistic attitude across outcomes.

AB - This paper develops a behavioral model of taxi driver passenger search behavior under uncertain conditions. We assume that, with a particular decision horizon in mind, a taxi driver chooses a particular strategy to search passengers. We differentiate between random search, maximum anticipated pick-up probability search, and maximum anticipated revenue search strategies. The probability of choosing a particular strategy is proportional to the weighted sum of the rewards associated with the possible outcomes of the decision, where the weights represent the driver’s beliefs, i.e. subjective probabilities, about the outcomes of the decision. Probability weightings are used to reweigh the belief set to capture drivers’ optimistic vs. pessimistic attitudes with respect to the uncertain outcomes. The reward function consists of two components: the anticipated monetary rewards of a decision within a certain time horizon, and the anticipated information value which captures the reduction in uncertainty as taxi drivers learn about their environment. The parameters of the value/reward function and the probability weighting are estimated using observed taxi trajectories derived from 1.5 million taxi global positioning system records. Results show that the probability weighting function signals an overall pessimistic attitude across outcomes.

KW - Behavioral model

KW - Discrete choice

KW - Information update

KW - Taxi drivers

KW - Uncertainty

UR - http://www.scopus.com/inward/record.url?scp=85092924581&partnerID=8YFLogxK

U2 - 10.1016/j.tbs.2020.09.008

DO - 10.1016/j.tbs.2020.09.008

M3 - Article

SN - 2214-367X

VL - 22

SP - 207

EP - 218

JO - Travel Behaviour and Society

JF - Travel Behaviour and Society

ER -