Crowdsourcing Regional Coverage Balancing Method Based on Transfer Learning in Taxi Service

With the in-depth study of taxi services, mobile crowdsourcing has become an emerging paradigm for solving location-based assignment tasks. However, only considering the task completion and regional coverage balance may reduce user’s satisfaction. Therefore, designing effective methods that not only address coverage balance but also take into account user’s satisfaction is a problem that needs to be addressed. We investigate the problem of regional coverage balance with users’ satisfaction and propose a service selection method based on transfer learning. This method comprises user’s trajectory prediction and service’s reputation to recommend services to users. In the user’s trajectory prediction part, an incentive mechanism is considered to ensure an even distribution of taxi services in each sub-region. Taking the service providers’ moving as the input, we design an adaptive ant colony algorithm that incorporates transfer learning to provide users with the optimal services. We validate the effectiveness of this method with datasets collected from the real world and compare the performance with existing regional coverage balance strategies. The experimental results show that user’s satisfaction increases by 53%, and the imbalance reduces by 24%.