In online-to-offline (O2O) on-demand services, customers place orders online and the O2O platform delivers products from stores to customers within a prescribed time. The platform usually hires crowd-sourced drivers as a cost-effective option owing to their flexibility. However, the delivery speed and delivery capacity of the crowd-sourced drivers vary considerably. This service inconsistency brings challenges in precisely matching the delivery supply and customer demand, which may significantly decrease the delivery efficiency. This study aims to address the challenges by proposing a personalized dispatch model, which integrates the order and driver's characteristics in the order assignment and routing decisions. To achieve this objective, two machine learning-based models are proposed to forecast the delivery speed of individual drivers in real time and customize their delivery capacity dynamically to develop a portrait of each driver's behaviour. Next, a personalized O2O order assignment and routing model is proposed with the integration of the two aforementioned models. We validate our model with a real dataset of one mainstream O2O platform in China. We run a comprehensive simulation to show the improvement in terms of on-time ratio and average delay time brought by the personalization of each characteristic, namely, delivery speed and delivery capacity. We then show that the proposed personalized model can reduce the average delay by 21.60% through comparison with actual routing decisions by the drivers,. The theoretical and numerical results shed light on the delivery management of the O2O on-demand services.
All Science Journal Classification (ASJC) codes
- Computer Science(all)
- Modeling and Simulation
- Management Science and Operations Research
- Information Systems and Management
- Machine learning