Staples, Inc.

Timothy Wee

Karthik Kumara

Courosh Mehanian

Tchavdar Dangaltchev

Ryan Applegate

Majid Hosseini

Jing Pan

Technology for predicting online user shopping behavior, such as whether a user will purchase a product, is described. An example method includes receiving current session data describing a current session for a current user, extracting a current clickstream from the current session data classifying the current clickstream as a purchase clickstream or a non-purchase clickstream by processing the current clickstream using one or more sets of Hidden Markov Model parameters produced by one or more Hidden Markov Models, and computing, using the one or more computing devices, a purchase probability that the current user will purchase a product during the current session based on the classifying.

Clickstream purchase prediction using Hidden Markov Models

The disclosure relates in some cases to a technology for selecting one or more promotions to be presented to online customers using Bayesian bandits and affinity-based dynamic user clustering In some embodiments, a computer-implemented method determines a set of offers is determined, and computes affinity scores measuring affinities of users to items included in the offers. The method builds an affinity score distribution for the offers and identifies clusters of affinity scores for the offers using the corresponding affinity score distribution.

Realtime feedback using affinity-based dynamic user clustering

A computer-implemented method and system are described for learning a vector representation for unique identification codes. An example method may include generating a unique identification code list using one or more virtual interaction contexts, the unique identification code list being a list of unique identification codes, selecting a target unique identification code in the unique identification code list, and determining, from the unique identification code list, an input set of unique identification codes using the target unique identification code, the input set including the target unique identification code and one or more context unique identification codes. Some implementations may further include inputting the input set of unique identification codes into a semantic neural network model, the semantic neural network model including one or more weight matrices, and modifying the one or more weight matrices using the input set of unique identification codes.

Learning a vector representation for unique identification codes

Technology for selecting promotion(s) to display in a page of an application for display to a user is described. An example method includes determining a promotion for a product; calculating for the promotion a posterior distribution of a user-action probability reflecting estimates for a user response to a display of the promotion for the product on a computing device of the user; determining the posterior distribution as collapsing beyond a certain threshold; responsive thereto, calculating an uncollapsed posterior distribution of the user-action probability reflecting modified estimates for the user response to the display of the promotion for the product on a computing device of the user; storing the uncollapsed posterior distribution of the user-action probability in a response database; and determining whether to select the promotion from the promotion database for display on a computing device of the user based on the modified estimates.

Promotion selection for online customers using Bayesian bandits

Technology for determining the order of the search results to maximize a financial goal is described. In an example embodiment, a method, implemented using the one or more computing devices, such as client and/or server devices, may receive a product search request from a user device associated with a user and retrieve a set of products from a product database based on the product search request. Based on a purchase probability and one or more of a margin and a price for that product, the method determines an expected financial gain for each of the products of the set and sorts the set of products into an ordered set of products having an order based on the expected financial gain associated with each of the products. The method may then provide the ordered set of products for display to the user on the user device.

Growing community of inventors

Burlingame, CA, United States of America

Timothy Wee