Predicting shopper supply during a time interval based on interactions by shoppers with a shopper assignment application during earlier time intervals

A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor of a computing system, cause the computing system to:

• identify a time interval for fulfillment of orders;
• determine a predicted number of shoppers available to fulfill orders during the identified time interval by applying a shopper supply model to information describing accesses to a shopper mobile application, from which shoppers select orders, by shoppers during one or more earlier time intervals, wherein the shopper supply model is a first neural network trained to predict a number of shoppers available to fulfill orders during any given time interval by:
• accessing first training examples, each first training example including an access pattern to a shopper mobile application by shoppers during a historical time interval labeled with a number of shoppers available to fulfill orders;
• applying parameters of the first neural network to the first training examples to predict numbers of shoppers available;
• and
• backpropagating to update the parameters of the first neural network to reduce errors between the predicted numbers of shoppers available to fulfill orders and the labeled numbers of shoppers available to fulfill orders;
• identify a target amount of time between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers;
• determine a shopper-order ratio value based on the predicted number of shoppers determined by the shopper supply model and a predicted number of orders for the identified time interval, the predicted number of orders being determined from numbers of orders received by the computing system for fulfillment during the one or more earlier time intervals;
• determine a predicted time for the identified time interval between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers by applying a selection prediction model to the determined shopper-order ratio value, wherein the selection prediction model is a second neural network trained to predict a time between presentation of orders to one or more shoppers and selection of one or more orders by the one or more shoppers for any given shopper-order ratio by:
• accessing second training examples, each second training example including a shopper-order ratio value labeled with a time interval between presentation of an order to one or more shoppers and selection of the order by the one or more shoppers;
• applying parameters of the second neural network to the second training examples to predict a time interval between presentation of orders to one or more shoppers and selection of one or more orders by the one or more shoppers;
• and
• backpropagating to update parameters of the second neural network to reduce errors between the predicted time interval and the labeled time interval;
• determine whether a supply gap exists during the identified time interval by comparing the predicted time for the identified time interval between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers to the target amount of time;
• adjust a number of shoppers allocated to fulfilling orders during the identified time interval based on the determination of whether a supply gap exists;
• and
• send the orders that are to be fulfilled during the identified time interval to client devices associated with the number of shoppers, causing the orders to be displayed at the client devices via the shopper mobile application associated with the number of shoppers.

A computing system comprising:

• a processor;
• and
• a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by the processor, cause the computing system to:
• identify a time interval among a plurality of time intervals in one or more days of a week for fulfillment of orders;
• determine a predicted number of shoppers available to fulfill orders during the identified time interval by applying a shopper supply model to information describing accesses to a shopper mobile application, from which shoppers select orders, by shoppers during one or more earlier time intervals, wherein the shopper supply model is a first neural network trained to predict a number of shoppers available to fulfill orders during any given time interval by:
• accessing first training examples, each first training example including an access pattern to a shopper mobile application by shoppers during a historical time interval labeled with a number of shoppers available to fulfill orders;
• applying parameters of the first neural network to the first training examples to predict numbers of shoppers available;
• and
• backpropagating to update the parameters of the first neural network to reduce errors between the predicted numbers of shoppers available to fulfill orders and the labeled numbers of shoppers available to fulfill orders;
• identify a target amount of time between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers;
• determine a shopper-order ratio value based on the predicted number of shoppers determined by the shopper supply model and a predicted number of orders for the identified time interval, the predicted number of orders being determined from numbers of orders received by the system for fulfillment during the one or more earlier time intervals;
• determine a predicted time for the identified time interval between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers by applying a selection prediction model to the determined shopper-order ratio value, wherein the selection prediction model is a second neural network trained to predict a time between presentation of orders to one or more shoppers and selection of one or more orders by the one or more shoppers for any given shopper-order ratio by:
• accessing second training examples, each second training example including a shopper-order ratio value labeled with a time interval between presentation of an order to one or more shoppers and selection of the order by the one or more shoppers;
• applying parameters of the second neural network to the second training examples to predict a time interval between presentation of orders to one or more shoppers and selection of one or more orders by the one or more shoppers;
• and
• backpropagating to update parameters of the second neural network to reduce errors between the predicted time interval and the labeled time interval;
• determine whether a supply gap exists during the identified time interval by comparing the predicted time for the identified time interval between presentation of orders to one or more shoppers and selection of one or more orders by one or more shoppers to the target amount of time;
• adjust a number of shoppers allocated to fulfilling orders during the identified time interval based on the determination of whether a supply gap exists;
• and
• send the orders that are to be fulfilled during the identified time interval to client devices associated with the number of shoppers, causing the orders to be displayed at the client devices via the shopper mobile application associated with the number of shoppers.