Owing to the shareability and spatial-temporal imbalance of free-floating bike-sharing (FFBS), the users may fail to pick up the bike at the desired location (i.e., the demand is truncated). Consequently, the demand partially migrates to nearby areas or is lost. Thus, the observed demand recorded in the system is not the underlying real demand. To address this issue, a framework for Demand Recovery considering demand Truncation, Migration, and spatial Correlation (DRTMC) was proposed. In detail, the real demand recovery is first modeled as a maximum-a-posteriori (MAP) problem. Then, the prior term and conditional term in MAP are formulated to jointly consider the demand truncation, migration, and spatial correlation. Finally, a tailored simulated annealing approach is applied to estimate the real demand. We then present the results using simulation data for validation and real-world data for empirical analysis. The validation results indicate that the demand recovered by the DRTMC model is much closer to the real demand value than the observed demand, and it performs better than all the benchmarks. For the case study, the FFBS data from Shanghai City in a prosperous area is chosen. It shows the distribution of observed demand is significantly different from that of real demand, emphasizing the importance of real demand recovery. The proposed DRTMC model enables researchers to capture the underlying real demand for FFBS and develop more effective rebalance strategies to improve FFBS service levels.
Recovering Real Demand for Free-Floating Bike-Sharing System Considering Demand Truncation, Migration, and Spatial Correlation
IEEE Transactions on Intelligent Transportation Systems ; 25 , 5 ; 4675-4691
2024-05-01
2312614 byte
Article (Journal)
Electronic Resource
English
Station-level Demand Prediction for Bike-Sharing System
IEEE | 2021
|A Method of Bike Sharing Demand Forecasting
Trans Tech Publications | 2014
|