Research
Current research projects and publications
Optimizing Order Assignments in On-Demand Delivery with Behavior Heterogeneity of Crowdsourced Drivers
Developed a Branch & Price algorithm integrated with XGBoost to predict driver behavior and optimize order assignments. The solution achieved a 62-71% reduction in delivery delays compared to traditional heuristics. This work addresses the stochastic nature of driver acceptance in gig-economy platforms.
Oversupply of Fissured Workforce and Operational Inefficiency: Evidence from Grab-Mode Competition in Food Delivery
Conducted a large-scale empirical analysis of 650,000+ delivery records. Utilized trajectory reconstruction and causal inference methods to quantify the impact of workforce competition on service quality. The model achieved an R² of 0.98, providing robust insights for platform policy design.
Modeling Liquidity and Survival: Inventory Speculation and Bankruptcy Risk Analysis
Formulated a Markov Decision Process (MDP) model to optimize inventory decisions under financial constraints. Implemented a CUDA-accelerated Dynamic Programming algorithm to solve the high-dimensional state space. Theoretically proved the log-concavity of the value function, ensuring convergence and stability.
Balancing Workforce Fissuring and Service Quality: Evidence from Dialysis Operations
Constructed a comprehensive longitudinal dataset (2015–2022) by integrating heterogeneous public databases (CMS, BLS, HCRIS); implemented a schema alignment algorithm based on feature similarity to resolve column inconsistencies across years. Analyzed the impact of outsourcing ratios on service quality, revealing a non-monotonic (inverted-U) relationship where a moderate proportion of contract nurses maximizes operational performance.