Dynamic worker selection and task allocation in crowdsourcing: A hybrid behavior-driven and supply-aware approach

Mobile crowdsourcing faces challenges from volatile worker behaviors and spatiotemporal resource imbalances. This paper proposes a hybrid behavior-driven and supply-aware framework for dynamic task allocation. Specifically, we develop a historical performance-driven dynamic credit scoring model that integrates long-term reputation with short-term completion status to quantify worker reliability. A nonlinear fitness transformation mechanism is then employed to map these behavioral metrics into definitive matching suitability, effectively amplifying high-quality worker-task pairings. Simultaneously, a dual-dynamic priority mechanism incorporating real-time supply–demand ratios is introduced to boost task utility in undersupplied areas, strategically guiding sensing capacity toward resource-scarce regions. To balance matching quality and efficiency under spatiotemporal constraints, an adaptive pruned two-stage strategy is employed. It implements reliability-based filtering to reduce computational overhead for regular tasks while ensuring global optimality for critical ones. Simulations on real-world datasets confirm the effectiveness of the proposed approach across varying scenarios.