OPTI: Order Preparation Time Inference for On-demand Delivery

On-demand delivery has become an increasingly popular urban service in recent years as it facilitates citizens’ daily lives significantly. In the fulfillment cycle, order preparation time estimation is extremely important and can be used for many applications, such improving dispatching estimation. Existing work generally based on high-cost physical devices or large-scale labeled training data, which are not feasible on-demand services. We solve this problem already collected different kinds of data from platform, e.g., courier’s reported arrival to merchant. Our intuition that couriers’ implicitly reflects time, leads a challenge: complicated correlations between time. To challenge, we design inference framework OPTI, first constructs self-supervised classification task infer coarse-grained then exploits semi-supervised learning transfer fine-grained inference. Experimental results show OPTI improve accuracy by 5% 17% compared state-of-the-art solutions.