US-12620014-B2 - System, method, and computer readable medium for determining attribute affinities for users

US12620014B2US 12620014 B2US12620014 B2US 12620014B2US-12620014-B2

Abstract

In some examples, a system may be configured to, for at least a first user of the plurality of users, implement a first set of operations that generate, for each of a first set of item types, attribute value data. Additionally, the system may implement a second set of operations that generate, for each of a second set of item types identified in catalogue data, clique data. Moreover, the system may, for the at least first user, implement a third set of operations that generate preference dependency data. Further, the system may, for the at least first user, based on the preference dependency data, the clique data, the attribute value data, generate, for each item type of a set of item types, output data including an affinity value for each item type of the first set of item types.

Inventors

Rahul Radhakrishnan Iyer
Shashank Kedia
Sushant Kumar
Kannan Achan

Assignees

WALMART APOLLO, LLC

Dates

Publication Date: 20260505
Application Date: 20211028

Claims (12)

1 . A system comprising: a database storing catalogue data and, for each user of a plurality of users of an online platform, transaction data, engagement data and search query data; at least one processor; and a memory resource storing instructions, that when executed by the at least one processor, causes the at least one processor to: for at least a first user of the plurality of users: based on the transaction data of the at least first user, implement a first set of operations that generate, for each of a first set of item types, attribute value data characterizing an attribute value for each attribute feature of one or more items associated with each of the first set of item types; apply a Naive Bayes model to the attribute value data to generate output data including an affinity value for each item type of the first set of item types, wherein the Naive Bayes model is a machine learning model stored in a machine learning database and trained based on attribute features that are treated as independent to each other; based on the catalogue data, implement a second set of operations that generate, for each of a second set of item types identified in the catalogue data, clique data characterizing one or more cliques based on graph clustering, each of the one or more cliques identifying a subset of attribute features of a set of attribute features that are mutually exclusive, wherein the second set of operations includes: based on the catalogue data, generating, for each item type identified from the catalogue data, a corresponding graph cluster comprising nodes and edges, wherein each node in the corresponding graph cluster represents an attribute feature of an item of the item type, wherein each edge in the corresponding graph cluster connects two nodes representing attribute features that occur concurrently, generating, for each item type identified from the catalogue data, a complementary graph cluster that is an inverse of the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes that are not connected by any edge in the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes representing attribute features that do not occur concurrently, and determining, based on the complementary graph cluster for each item type identified from the catalogue data, the set of attribute features that are mutually exclusive; create a first training dataset that comprises the clique data and attribute features including the set of attribute features that are mutually exclusive; re-train the Naive Bayes model using the first training dataset to generate a first Bayesian model that accounts for a mutual exclusivity between attribute features of items; based at least in part on the transaction data, engagement data and search query data of at least the first user, implement a third set of operations that generate dependency data characterizing one or more dependencies between each attribute feature of a set of predetermined attribute features based on at least one predetermined dependency template; create a second training dataset that comprises the first training dataset, the dependency data, and the set of predetermined attribute features; re-train the first Bayesian model using the second training dataset to generate a second Bayesian model that accounts for both the mutual exclusivity between attribute features of items and a dependency relationship between attribute features of items; and apply the second Bayesian model to the attribute value data to generate updated output data.
2 . The system of claim 1 , wherein the first set of operations includes: based on the transaction data of at least the first user, identifying the first set of item types; based on the transaction data, generating, for each item type of the first set of item types, attribute feature data, the attribute feature data characterizing a set of attribute features; based on the transaction data, the engagement data and set of user attribute features, generate attribute value data that indicates, for each of the set of attribute features, the attribute value.
3 . The system of claim 1 , wherein the affinity value of each item type of the set of item types indicates a likelihood of an occurrence of a purchase event between the at least first user and one or more items of the corresponding item type.
4 . The system of claim 1 , wherein each attribute feature is associated with a preference of a preference set, the set of preferences including at least one of (i) type preferences for various products, (ii) price sensitivity at a product/product-type level, (iii), brand sensitivity and preferences, (iv) restriction preferences, (v) restricted foods preferences, (vi) dietary methods preferences, (vii) dietary needs preferences, (viii) allergens preferences, (viv) container types preferences, and (x) quantity preferences.
5 . The system of claim 1 , wherein the attribute value data characterizes a likelihood of an occurrence of a purchase event between the at least first user and a particular item of a particular item type with the corresponding attribute feature.
6 . The system of claim 1 , wherein the clique data is further based on search query data.
7 . A computer-implemented method comprising: for at least a first user of a plurality of users: based on transaction data of at least the first user, implementing a first set of operations that generate, for each of a first set of item types, attribute value data characterizing an attribute value for each attribute feature of one or more items associated with each of the first set of item types; applying a Naïve Bayes model to the attribute value data to generate output data including an affinity value for each item type of the first set of item types, wherein the Naive Bayes model is a machine learning model stored in a machine learning database and trained based on attribute features that are treated as independent to each other; based on catalogue data, implementing a second set of operations that generate, for each of a second set of item types identified in the catalogue data, clique data characterizing one or more cliques based on graph clustering, each of the one or more cliques identifying a subset of attribute features of a set of attribute features that are mutually exclusive, wherein the second set of operations includes: based on the catalogue data, generating, for each item type identified from the catalogue data, a corresponding graph cluster comprising nodes and edges, wherein each node in the corresponding graph cluster represents an attribute feature of an item of the item type, wherein each edge in the corresponding graph cluster connects two nodes representing attribute features that occur concurrently, generating, for each item type identified from the catalogue data, a complementary graph cluster that is an inverse of the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes that are not connected by any edge in the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes representing attribute features that do not occur concurrently, and determining, based on the complementary graph cluster for each item type identified from the catalogue data, the set of attribute features that are mutually exclusive; creating a first training dataset that comprises the clique data and attribute features including the set of attribute features that are mutually exclusive; re-training the Naive Bayes model using the first training dataset to generate a first Bayesian model that accounts for a mutual exclusivity between attribute features of items; based at least in part on the transaction data, engagement data and search query data of at least the first user, implementing a third set of operations that generate dependency data characterizing one or more dependencies between each attribute feature of a set of predetermined attribute features based on at least one predetermined dependency template; creating a second training dataset that comprises the first training dataset, the dependency data, and the set of predetermined attribute features; re-training the first Bayesian model using the second training dataset to generate a second-neural network Bayesian model that accounts for both the mutual exclusivity between attribute features of items and a dependency relationship between attribute features of items; and applying the second Bayesian model to the attribute value data to generate updated output data.
8 . The computer-implemented method of claim 7 , wherein the first set of operations includes: based on the transaction data of at least the first user, identifying the first set of item types; based on the transaction data, generating, for each item type of the first set of item types, attribute feature data, the attribute feature data characterizing a set of attribute features; based on the transaction data, the engagement data and set of user attribute features, generate attribute value data that indicates, for each of the set of attribute features, the attribute value.
9 . The computer-implemented method of claim 7 , wherein the affinity value of each item type of the set of item types indicates a likelihood of an occurrence of a purchase event between the at least first user and one or more items of the corresponding item type.
10 . The computer-implemented method of claim 7 , wherein each attribute feature is associated with a preference of a preference set, the set of preferences including at least one of (i) type preferences for various products, (ii) price Sensitivity at a product/product-type level, (iii), brand Sensitivity and Preferences, (iv) Restriction preferences, (v) Restricted Foods Preferences, (vi) Dietary Methods Preferences, (vii) Dietary Needs Preferences, (viii) Allergens Preferences, (viv) Container Types Preferences, and (x) Quantity Preferences.
11 . The computer-implemented method of claim 7 , wherein the attribute value data characterizes a likelihood of an occurrence of a purchase event between the at least first user and a particular item of a particular item type with the corresponding attribute feature.
12 . A non-transitory computer readable medium storing instructions, that when executed by at least one processor, causes a system to: for at least a first user of a plurality of users: based on transaction data of at least the first user, implement a first set of operations that generate, for each of a first set of item types, attribute value data characterizing an attribute value for each attribute feature of one or more items associated with each of the first set of item types; apply a Naive Bayes model to the attribute value data to generate output data including an affinity value for each item type of the first set of item types, wherein the Naive Bayes model is a machine learning model stored in a machine learning database and trained based on attribute features that are treated as independent to each other; based on catalogue data, implement a second set of operations that generate, for each of a second set of item types identified in the catalogue data, clique data characterizing one or more cliques based on graph clustering, each of the one or more cliques identifying a subset of attribute features of a set of attribute features that are mutually exclusive, wherein the second set of operations includes: based on the catalogue data, generating, for each item type identified from the catalogue data, a corresponding graph cluster comprising nodes and edges, wherein each node in the corresponding graph cluster represents an attribute feature of an item of the item type, wherein each edge in the corresponding graph cluster connects two nodes representing attribute features that occur concurrently, generating, for each item type identified from the catalogue data, a complementary graph cluster that is an inverse of the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes that are not connected by any edge in the corresponding graph cluster, wherein each edge in the complementary graph cluster connects two nodes representing attribute features that do not occur concurrently, and determining, based on the complementary graph cluster for each item type identified from the catalogue data, the set of attribute features that are mutually exclusive; create a first training dataset that comprises the clique data and attribute features including the set of attribute features that are mutually exclusive; re-train the Naive Bayes model using the first training dataset to generate a first Bayesian model that accounts for a mutual exclusivity between attribute features of items; based at least in part on the transaction data, engagement data and search query data of at least the first user, implement a third set of operations that generate dependency data characterizing one or more dependencies between each attribute feature of the set of predetermined attribute features based on at least one predetermined dependency template; create a second training dataset that comprises the first training dataset, the dependency data, and the set of predetermined attribute features; re-train the first Bayesian model using the second training dataset to generate a second Bayesian model that accounts for both the mutual exclusivity between attribute features of items and a dependency relationship between attribute features of items; and apply the second Bayesian model to the attribute value data to generate updated output data.

Description

TECHNICAL FIELD The disclosure relates generally to network services, and more specifically, to automatically determining, generating, and providing user profiles. BACKGROUND At least some ecommerce entities can include recommendation systems that can personalize a user experience for different channels, such as an ecommerce platform with a search engine system, provided by the ecommerce entity. Conventionally, the recommendation systems of an ecommerce entities are based on information extracted from catalogue data established by the ecommerce entity. Such systems, may not take into account particular attributes of a user the recommendation is making recommendations for, and as such, may affect the recommendation-to-purchase rates (e.g., the rate at which users purchase items or products appearing in recommendations provided by the recommendation system) and may result in wasted computational resources in providing such recommendations (e.g., the recommendations are ignored). For example, an ecommerce entity may provide an online ecommerce platform, such as a website, along with a search engine system that can enable customer to search for products that the ecommerce entity provides. The website may include a search bar that allows the users to enter search terms, such as one or more words, that the website uses to search for products. In response to the search terms, a recommendation system of the website, may implement a search algorithm to generate a search result including products that meet the requirements of the search algorithm. However, such requirements may be based on information extracted from catalogue data established by the retailer, and may not take into account particular attributes of the user. As such, in such conventional systems, these search results may not be as accurate to the user submitting them and may affect the search-to-purchase conversion rate of the online ecommerce platform (e.g., the rate at which users purchase items or products appearing in search results stemming from search queries requested by the users). SUMMARY The embodiments described herein are directed to recommendation systems associated with one or more channels of ecommerce entities, such as an online ecommerce platform with a search engine. As described herein a channel of the one or more channels may refer to a particular data process or pipeline, such as those associated with as search engine system of an online ecommerce platform of an ecommerce entity, a notification system, and/or a complementary item recommendation system. In accordance with various embodiments, exemplary computing systems may be implemented in any suitable hardware or hardware and software, such as in any suitable computing device. In some embodiments a system may include a database storing catalogue data and, for each user of a plurality of users of an online platform, transaction data, engagement data and search query data. Additionally, the system may include at least one processor executes the instructions to, for at least a first user of the plurality of users and based on the transaction data of the at least first user, implement a first set of operations that generate, for each of a first set of item types, attribute value data characterizing an attribute value for each attribute feature of one or more items associated with each of the first set of item types. In some examples, each attribute feature being associated with preference of a set of preferences. Moreover, the at least one processor executes the instructions to, for the at least first user and based on the catalogue data, implement a second set of operations that generate, for each of a second set of item types identified in the catalogue data, clique data characterizing one or more cliques. In some examples, each of the one or more cliques identifying a subset of attribute features of a set of attribute features that are mutually exclusive. Further, the at least one processor executes the instructions to, for the at least first user and based at least in part on the transaction data, engagement data and search query data of at least the first user, implement a third set of operations that generate dependency data characterizing one or more dependencies between each preference of the set of preferences. Additionally, the at least one processor executes the instruction to, for the at least first user and based on the dependency data, the clique data, the attribute value data, generate, for each item type of a set of item types, output data including an affinity value for each item type of the first set of item types. In other embodiments, a computer-implemented method is provided that includes, for at least a first user of the plurality of users and based on the transaction data of the at least first user, implementing a first set of operations that generate, for each of a first set of item types, attribute value data characterizing an attribute value for each attribu