Using a machine learning model to predict a user's quantity ceiling for different categories of items in a catalog

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

• receiving, at a computer system, an order from a user, the order including one or more items, each item having a corresponding item category;
• retrieving one or more previously fulfilled orders for the user that the computer system previously fulfilled;
• identifying a ceiling of the user for an item category by applying a ceiling prediction model to information describing the order, the information describing the order including quantities of items included in the order and to the one or more previously fulfilled orders, the ceiling of the user for the item category comprising a maximum quantity of items from the item category likely to be included in the order, wherein the ceiling prediction model comprises a machine-learning model trained by:
• obtaining a training dataset including a plurality of training examples, each training example including, a training item category, a training history of previously fulfilled orders, information describing a training order, and contextual information about the training order, each training example having a label indicating a ceiling for the training item category;
• applying the ceiling prediction model to each training example of the training dataset to generate a predicted ceiling for the training item category;
• scoring the ceiling prediction model using a loss function and the label of the training example;
• and
• updating one or more parameters of the ceiling prediction model by backpropagation based on the scoring until one or more criteria are satisfied;
• selecting content for presentation by the computer system to the user based on the ceiling of the user for the item category;
• and
• sending the selected content to a device of the user, the sending causing the device of the user to display the selected content.

A system comprising:

• a processor;
• and
• a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by the processor, cause the processor to perform steps comprising:
• receiving, at a computer system, an order from a user, the order including one or more items, each item having a corresponding item category;
• retrieving one or more previously fulfilled orders for the user that the computer system previously fulfilled;
• identifying a ceiling of the user for an item category by applying a ceiling prediction model to information describing the order, the information describing the order including quantities of items included in the order and to the one or more previously fulfilled orders, the ceiling of the user for the item category comprising a maximum quantity of items from the item category likely to be included in the order, wherein the ceiling prediction model comprises a machine-learning model trained by:
• obtaining a training dataset including a plurality of training examples, each training example including, a training item category, a training history of previously fulfilled orders, information describing a training order, and contextual information about the training order, each training example having a label indicating a ceiling for the training item category;
• applying the ceiling prediction model to each training example of the training dataset to generate a predicted ceiling for the training item category;
• scoring the ceiling prediction model using a loss function and the label of the training example;
• and
• updating one or more parameters of the ceiling prediction model by backpropagation based on the scoring until one or more criteria are satisfied;
• selecting content for presentation by the computer system to the user based on the ceiling of the user for the item category;
• and
• sending the selected content to a device of the user, the sending causing the device of the user to display the selected content.