3-9: Dynamic Prediction of Shared Micromobility Usage with Multi-Task Learning
Principal Investigator | Yu Zhang, Ph.D. |
Final Report (DOI) | Available Soon |
TRID | Available Soon |
Policy Brief | Available Soon |
RIP | Available Soon |
Abstract
Shared micromobility can not only replace short-distance vehicle trips, but also solve first/last-mile problem of public transit by leveraging the flexibility and efficient door-to-door accessibility of the system. Shared micromobility programs have been embraced by many cities in the US and worldwide. How to improve the effectiveness, safety, and equity of the programs, however, is the area that deserves further investigation. Supply-demand imbalance is an intrinsic feature of the programs, which hinders the realization of maximal benefit of the programs. Thus, this proposed study will predict the usage of different micro-vehicles in a comprehensive shared micromobility program by applying a multi-task learning (MTL) method. Hard-parameter and soft-parameter models of MTL will be developed considering the interactions of different types micro-vehicles. The models will be tested with a case study, shared micromobility program in the City of Tampa, to compare the performance of the models. In addition, by using the outcomes of the learning methods, incentive programs for encouraging operators to respond to the imbalance of supply and demand will be explored, and research directions to further improving shared micromobility will be discussed.