Supply and demand aware synthetic data generation for on-demand traffic with real-world characteristics

The on-demand economy has attracted significant attention in recent years, with a rapid growth in on-demand services ranging from ride-hailing to package delivery and grocery pickup. However, real-world spatio-temporal data that can be used for evaluating research on on-demand brokers design and supply-demand regulation are either not publicly available or are very limited in their spatial coverage. Research efforts in generating synthetic spatio-temporal datasets such as traffic generators have only focused on one side of the business model, particularly the demand side, and thus are not convenient for studying market variations such as the problem of supply-demand imbalance. In addition, many of these generators do not accurately reflect real-world data characteristics. In this paper, we propose a supply and demand aware framework for generating synthetic datasets for the purpose of designing on-demand spatial service brokers, while also capturing real-world data characteristics by leveraging multiple publicly available data sources. We also present an evaluation of the quality and performance of our proposed framework.