Abstract: Features related to a system and method for scheduling a resources to perform discrete tasks are described. The scheduling features include generating schedules predicted to appeal to the tasked resource (e.g., delivery partner) such as by time, day of the week, location, item types, etc. Using machine learning, the schedule and terms thereof can be dynamically generated to suit the tastes of each tasked resource and the overall demand for services. Using historical data, the modeling also accounts for likelihood an offer will be accepted and risk of cancellation for a given resource. The machine learning may be based on a mixed integer problem as constrained by partner and system capacity parameters.

Abstract: Embodiments herein describe a predictive delivery planning system that includes a forecaster that predicts simulated orders (e.g., forecasted orders) for multiple different demand scenarios. Once the simulated orders are selected, a route planner can generate routes for delivering the simulated and actual customer orders for each scenario. The planning system then converts these routes in labor plans indicating the amount of time a delivery driver would need to deliver the orders. The planning system identifies a set of labor blocks from the labor plans and determines whether these blocks satisfy a utilization threshold. Put differently, the planning system uses a releasing policy that releases labor blocks whose expected utilization is higher than a predetermined threshold. The released labor blocks are then displayed to delivery drivers who can then select how many of the labor blocks they would like to work.

Abstract: Features related to a system and method for scheduling a resources to perform discrete tasks are described. The scheduling features include generating schedules predicted to appeal to the tasked resource (e.g., delivery partner) such as by time, day of the week, location, item types, etc. Using machine learning, the schedule and terms thereof can be dynamically generated to suit the tastes of each tasked resource and the overall demand for services. Using historical data, the modeling also accounts for likelihood an offer will be accepted and risk of cancellation for a given resource. The machine learning may be based on a mixed integer problem as constrained by partner and system capacity parameters.

Abstract: Features related to a system and method for scheduling a resources to perform discrete tasks are described. The scheduling features include generating schedules predicted to appeal to the tasked resource (e.g., delivery partner) such as by time, day of the week, location, item types, etc. Using machine learning, the schedule and terms thereof can be dynamically generated to suit the tastes of each tasked resource and the overall demand for services. Using historical data, the modeling also accounts for likelihood an offer will be accepted and risk of cancellation for a given resource. The machine learning may be based on a mixed integer problem as constrained by partner and system capacity parameters.