US 12,475,411 B2
Using computer model to determine availability of service option for delivery of order placed with online system
Zi Wang, Mountain View, CA (US); Houtao Deng, Sunnyvale, CA (US); Xiangyu Wang, San Jose, CA (US); Ganesh Krishnan, San Francisco, CA (US); and Aman Jain, Barrie (CA)
Assigned to Maplebear Inc., San Francisco, CA (US)
Filed by Maplebear Inc., San Francisco, CA (US)
Filed on Jun. 16, 2023, as Appl. No. 18/210,976.
Prior Publication US 2024/0420037 A1, Dec. 19, 2024
Int. Cl. G06Q 10/04 (2023.01); G06Q 10/083 (2024.01); G06Q 30/0601 (2023.01)
CPC G06Q 10/04 (2013.01) [G06Q 10/083 (2013.01); G06Q 30/0601 (2013.01)] 19 Claims
OG exemplary drawing
 
1. A method comprising, at a computer system comprising a processor and a computer-readable medium:
receiving, via a network and from a device associated with a user of an online system, order data including information about a list of items in an order placed by the user via a user interface of the device associated with the user, information about a delivery location for the order, and a timeframe during which items from the list should be delivered to the delivery location;
receiving, via the network, supply data including information about a current level of available supply at the online system for servicing orders;
responsive to receiving the order data, accessing a machine-learning model of the online system, wherein the machine-learning model includes a multilayer neural network and is trained to predict a time to accept servicing of the order by the online system, training of the machine-learning model comprises:
generating a set of training examples including data related to a collection of past orders,
obtaining a label for each training example from the set of training examples,
applying a set of parameters of the machine-learning model to the set of training examples to generate an output for each training example from the set of training examples, the set of parameters including weights and biases that are applied at each neuron of the multilayer neural network,
comparing the output for each training example to the label using a loss function to generate a score for each training example, and
updating, through a back-propagation process, the set of parameters of the machine-learning model using the score for each training example;
applying the machine-learning model to the order data and the supply data to generate a value of a metric for the order that is indicative of the time to accept servicing of the order;
comparing the value of the metric to a first threshold value;
determining, based on a result of comparing the value of the metric to the first threshold value, a service option of a plurality of service options for servicing the order;
causing the device associated with the user to display the user interface with an availability of the service option for servicing the order;
generating, over a time period, training order data including information about a plurality of values of the metric indicative of a plurality of times to accept servicing of a set of orders, wherein the plurality of times to accept are generated by the machine-learning model, and the plurality of times to accept include the time to accept;
collecting, over the time period, information about levels of available supply at the online system for servicing the set of orders; and
re-training the machine-learning model by updating, through the back-propagation process, the set of parameters of the machine-learning model using the training order data and the information about levels of available supply.