Data-Driven Capacity Management of Ride-Hailing Service with Prescriptive Analytics and Deep Learning

This paper studies the capacity management of ride-hailing services with unstable passenger demand. In the existing parametric approaches, normally, a two-stage prediction-then-optimization (PTO) paradigm is used to implement demand prediction and capacity management sequentially, which generally achieves suboptimal performance because of the information loss in demand prediction. To attain optimality, in this study, we integrate these two stages into a prediction and optimization (P&O) paradigm and formulate a deep learning approach consisting of a weighted sample average approximation (WSAA) module and VMD-CNN-BiLSTM-AM (V-CBA) module (V-CBA integrates variational mode decomposition (VMD), convolutional neural network (CNN), bidirectional long short-term memory (BiLSTM), and attention mechanism (AM)), where the WSAA module embeds with k-nearest neighbors, kernel regression, or a decision tree to select significant historical samples, and the V-CBA module couples decomposition, convolution, recursion, attention, and optimization together for complex nonlinear mapping. Based on real-world data sets, the deep learning-based P&O paradigm is demonstrated to have superior performance in capacity management of ride-hailing services.

History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning.

Funding: This work was supported by the National Natural Science Foundation of China [Grants W2411066, U2469202, 72401019, and 72401018].

Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information (https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2024.0595) as well as from the IJOC GitHub software repository (https://github.com/INFORMSJoC/2024.0595). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/.