DoM

In the era of ubiquitous geographic information, generating Origin–Destination (OD) flows within urban unseen regions remains a critical challenge for mitigating data scarcity. Current approaches face significant trade-offs. Data-driven methods, while excelling in nonlinear fitting, rely heavily on extensive training data, resulting in limited cross-region generalizability. Mechanism-driven methods alleviate data dependence by incorporating established theories of human mobility, but often struggle to capture complex, nonlinear human behaviors. To address these issues, we propose the knowledge-guided Deep Opportunity Model (DoM).