2021/04/06 by Haritha Jayasinghe, Jayasinghe, Haritha, Tarindu Jayatilaka +5
Business, Management and Accounting · Engineering · #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Machine Learning (cs.LG) #Sharing Economy and Platforms #Transportation and Mobility Innovations
paper · pdf · doi:10.48550/arxiv.2104.02661
openalex publication_date 2021/04/06 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
The rapid growth of ride-hailing platforms has created a highly competitive\nmarket where businesses struggle to make profits, demanding the need for better\noperational strategies. However, real-world experiments are risky and expensive\nfor these platforms as they deal with millions of users daily. Thus, a need\narises for a simulated environment where they can predict users' reactions to\nchanges in the platform-specific parameters such as trip fares and incentives.\nBuilding such a simulation is challenging, as these platforms exist within\ndynamic environments where thousands of users regularly interact with one\nanother. This paper presents a framework to mimic and predict user,\nspecifically driver, behaviors in ride-hailing services. We use a data-driven\nhybrid reinforcement learning and imitation learning approach for this. First,\nthe agent utilizes behavioral cloning to mimic driver behavior using a\nreal-world data set. Next, reinforcement learning is applied on top of the\npre-trained agents in a simulated environment, to allow them to adapt to\nchanges in the platform. Our framework provides an ideal playground for\nride-hailing platforms to experiment with platform-specific parameters to\npredict drivers' behavioral patterns.\n