US-12620015-B2 - System, non-transitory computer readable medum, and method for determining temporal loyalty

US12620015B2US 12620015 B2US12620015 B2US 12620015B2US-12620015-B2

Abstract

Systems and methods for attribute recommendation are disclosed. Transaction data related a user is received and attribute recommendations for the user are generated based on the transaction data. The attribute recommendations are generated by a variational inference model configured using a transaction matrix and a loyalty matrix. A set of N recommendations is generated by ranking the generated attribute recommendations based on a combined transaction score and loyalty score and a user interface is generated including the set of N recommendations.

Inventors

Venugopal Mani
Ramasubramanian Balasubramanian
Sushant Kumar
Kannan Achan
Abhinav Mathur

Assignees

WALMART APOLLO, LLC

Dates

Publication Date: 20260505
Application Date: 20220901

Claims (20)

1 . A system, comprising: a processor; and a non-transitory memory storing instructions that, when executed, cause the processor to: receive, via a network interface, transaction data related a user; generate, based on the transaction data, a transaction matrix characterizing user interactions with each of a plurality of attributes and a loyalty matrix characterizing a number of the user interactions associated with each attribute of the plurality of attributes, wherein each number of the user interactions is weighted based on an elapsed time since each corresponding user interaction; generate, by a modified collaborative filtering model, a distribution of the transaction matrix and a distribution of the loyalty matrix, wherein the modified collaborative filtering model comprises a set of parameters that are used for the generation of each of the distribution of the transaction matrix and the distribution of the loyalty matrix, wherein the modified collaborative filtering mode is trained based on a set of training data comprising transaction matrices and loyalty matrices, and wherein at least a portion of the set of parameters are determined based on applying a variational inference to a generated distribution of the transaction matrices and a generated distribution of the loyalty matrices to determine a least divergence from a true distribution of each of the transaction matrices and the loyalty matrices; generate a transaction entry for each of one or more user-attribute pairs based on the distribution of the transaction matrix; generate a temporal entry for each of the one or more user-attribute pairs based on the distribution of the loyalty matrix; determine an overall score for the one or more user-attribute pairs by combining the transaction entry and the temporal entry for each of the one or more user-attribute pairs; generate attribute recommendations for the user based on the transaction data, wherein the attribute recommendations are generated by a variational inference model that receives the transaction matrix and the loyalty matrix; generate a set of N recommendations by ranking the generated attribute recommendations based on the overall score of each user-attribute pair; generate a user interface including the set of N recommendations; and transmit, via the network interface, the user interface for display to the user.
2 . The system of claim 1 , wherein the variational inference model includes a variational distribution which is a proxy-posterior that is least-divergent from a true posterior p(θ|T,L,), where θ represents latent variables T is the transaction matrix, and L is the loyalty matrix.
3 . The system of claim 1 , wherein the attribute recommendations are generated by a posterior predictive function generated by the variational inference model.
4 . The system of claim 3 , wherein the posterior predictive function uses a likelihood function P(T,L|θ,H), where θ represents latent variables, T is the transaction matrix, L is the loyalty matrix, and H represents a set of hyperparameters.
5 . The system of claim 1 , wherein the combined transaction score and loyalty score are generated by combining transaction and loyalty distributions.
6 . The system of claim 5 , wherein the transaction and loyalty distributions are calculated as: T p ⁢ q ∼ N ⁡ ( K t ( U P T ⁢ V q + b ⁢ u p ) + φ T , ( γ ⁢ B ) - 1 ) L p ⁢ q ∼ N ⁡ ( K l ( U P T ⁢ V q + b ⁢ u p ) + φ l , ( ( 1 - γ ) ⁢ B ) - 1 ) where Tpq is a transaction entry for a user p and an attribute q, Lpq is a temporal loyalty entry for the user p and the attribute q, U P T and V q are embedding vectors, bu p is a bias vector, and φ l and γ are hyperparameters.
7 . The system of claim 1 , wherein initial embedding values implemented by the variational inference model are provided by a graphical data set.
8 . The system of claim 7 , wherein the graphical data set includes a heterogeneous user interaction graph (G) defined as: G =( V,E,T ) where V is a set of vertices, E is a set of edges, and Tis a set of vertex types.
9 . A non-transitory computer readable medium having instructions stored thereon, wherein the instructions, when executed by a processor cause a device to perform operations comprising: receiving, via a network interface, transaction data related a user; generating, based on the transaction data, a transaction matrix characterizing user interactions with each of a plurality of attributes and a loyalty matrix characterizing a number of the user interactions associated with each attribute of the plurality of attributes, wherein each number of the user interactions is weighted based on an elapsed time since each corresponding user interaction; generating, by a modified collaborative filtering model, a distribution of the transaction matrix and a distribution of the loyalty matrix, wherein the modified collaborative filtering model comprises a set of parameters that are used for the generation of each of the distribution of the transaction matrix and the distribution of the loyalty matrix, wherein the modified collaborative filtering mode is trained based on a set of training data comprising transaction matrices and loyalty matrices, and wherein at least a portion of the set of parameters are determined based on applying a variational inference to a generated distribution of the transaction matrices and a generated distribution of the loyalty matrices to determine a least divergence from a true distribution of each of the transaction matrices and the loyalty matrices; generating a transaction entry for each of one or more user-attribute pairs based on the distribution of the transaction matrix; generating a temporal entry for each of the one or more user-attribute pairs based on the distribution of the loyalty matrix; determining an overall score for the one or more user-attribute pairs by combining the transaction entry and the temporal entry for each of the one or more user-attribute pairs; generating attribute recommendations for the user based on the transaction data, wherein the attribute recommendations are generated by a variational inference model that receives the transaction matrix and the loyalty matrix; generating a set of N recommendations by ranking the generated attribute recommendations based on the overall score of each user-attribute pair; generating a user interface including the set of N recommendations; and transmitting, via the network interface, the user interface for display to the user.
10 . The non-transitory computer readable medium of claim 9 , wherein the variational inference model includes a variational distribution which is a proxy-posterior that is least-divergent from a true posterior p(θ|T,L,), where θ represents latent variables, T is the transaction matrix, and L is the loyalty matrix.
11 . The non-transitory computer readable medium memory of claim 9 , wherein the attribute recommendations are generated by a posterior predictive function generated by the variational inference model.
12 . The non-transitory computer readable medium of claim 11 , wherein the posterior predictive function uses a likelihood function P(T,L|θ,H), where θ represents latent variables, T is the transaction matrix, L is the loyalty matrix, and H represents a set of hyperparameters.
13 . The non-transitory computer readable medium of claim 9 , wherein the combined transaction score and loyalty score are generated by combining transaction and loyalty distributions.
14 . The non-transitory computer readable medium of claim 13 , wherein the transaction and loyalty distributions are calculated as: T p ⁢ q ∼ N ⁡ ( K t ( U P T ⁢ V q + b ⁢ u p ) + φ T , ( γ ⁢ B ) - 1 ) L p ⁢ q ∼ N ⁡ ( K l ( U P T ⁢ V q + b ⁢ u p ) + φ l , ( ( 1 - γ ) ⁢ B ) - 1 ) where Tpq is a transaction entry for a user p and an attribute q, Lpq is a temporal loyalty entry for the user p and the attribute q, U P T and V q are embedding vectors, bu p is a bias vector, and φ l and γ are hyperparameters.
15 . The non-transitory computer readable medium of claim 9 , wherein initial embedding values implemented by the variational inference model are provided by a graphical data set.
16 . The non-transitory computer readable medium of claim 15 , wherein the graphical data set includes a heterogeneous user interaction graph (G) defined as: G =( V,E,T ) where V is a set of vertices, E is a set of edges, and T is a set of vertex types.
17 . A method by at least one processor, the method comprising: receiving, via a network interface, transaction data related a user; generating, based on the transaction data, a transaction matrix characterizing user interactions with each of a plurality of attributes and a loyalty matrix characterizing a number of the user interactions associated with each attribute of the plurality of attributes, wherein each number of the user interactions is weighted based on an elapsed time since each corresponding user interaction; generating, by a modified collaborative filtering model, a distribution of the transaction matrix and a distribution of the loyalty matrix, wherein the modified collaborative filtering model comprises a set of parameters that are used for the generation of each of the distribution of the transaction matrix and the distribution of the loyalty matrix, wherein the modified collaborative filtering mode is trained based on a set of training data comprising transaction matrices and loyalty matrices, and wherein at least a portion of the set of parameters are determined based on applying a variational inference to a generated distribution of the transaction matrices and a generated distribution of the loyalty matrices to determine a least divergence from a true distribution of each of the transaction matrices and the loyalty matrices; generating a transaction entry for each of one or more user-attribute pairs based on the distribution of the transaction matrix; generating a temporal entry for each of the one or more user-attribute pairs based on the distribution of the loyalty matrix; determining an overall score for the one or more user-attribute pairs by combining the transaction entry and the temporal entry for each of the one or more user-attribute pairs; generating attribute recommendations for the user based on the transaction data, wherein the attribute recommendations are generated by a variational inference model that receives the transaction matrix and the loyalty matrix; generating a set of N recommendations by ranking the generated attribute recommendations based on the overall score of each user-attribute pair; generating a user interface including the set of N recommendations; and transmitting, via the network interface, the user interface for display to the user.
18 . The method of claim 17 , wherein the variational inference model includes a variational distribution which is a proxy-posterior that is least-divergent from a true posterior p(θ|T,L), where θ represents latent variables, T is the transaction matrix, and L is the loyalty matrix.
19 . The method of claim 17 , wherein the attribute recommendations are generated by a posterior predictive function generated by the variational inference model, and wherein the posterior predictive function uses a likelihood function P(T,L|θ,H), where θ represents latent variables, T is the transaction matrix, L is the loyalty matrix, and H represents a set of hyperparameters.
20 . The method of claim 17 , wherein initial embedding values implemented by the variational inference model are provided by a graphical data set.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application claims benefit under 35 U.S.C. 119(e) to U.S. Provisional Patent Appl. No. 63/264,925, filed Dec. 3, 2021, entitled “System and Methods for Determining Temporal Loyalty,” the disclosure of which is incorporated herein by reference in its entirety. TECHNICAL FIELD This application relates generally to attribute recommendation and, more particularly, to attribute recommendation using a time-related matrix. BACKGROUND Interactions between users and systems may vary over time, with certain user interactions or preferences becoming more or less prevalent over predetermined time periods. In environments configured to provide interactions with multiple variations of similar elements, users may develop or change preferences for interactions with specific versions of a variant element or item. For example, in an e-commerce environment, an individual user may express one or more preferences for brands, styles, etc. of specific items available within the e-commerce interface. These preferences may change over time. Personalizing user interfaces, such as e-commerce interfaces, drives user satisfaction and engagement for network interfaces. Current network interfaces are capable of suggesting categories or items based on prior interactions but fail to account for user preferences that can be identified based on temporal interactions with the system. Current systems implement a graphical structure that ignores or excludes temporal data from the consideration. SUMMARY In various embodiments, a system is disclosed. The system includes a non-transitory memory having instructions stored thereon and a processor configured to read the instructions. The processor is configured to receive transaction data related a user, generate attribute recommendations for the user based on the transaction data, generate a set of N recommendations by ranking the generated attribute recommendations based on a combined transaction score and loyalty score, and generate a user interface including the set of N recommendations. The attribute recommendations are generated by a variational inference model configured using a transaction matrix and a loyalty matrix. In various embodiments, a non-transitory computer readable medium having instructions stored thereon is disclosed. The instructions, when executed by a processor cause a device to perform operations including receiving transaction data related a user, generating attribute recommendations for the user based on the transaction data, generating a set of N recommendations by ranking the generated attribute recommendations based on a combined transaction score and loyalty score, and generating a user interface including the set of N recommendations. The attribute recommendations are generated by a variational inference model configured using a transaction matrix and a loyalty matrix. In various embodiments, a computer-implemented method is disclosed. The computer-implemented method includes the steps of receiving transaction data related a user, generating attribute recommendations for the user based on the transaction data, generating a set of N recommendations by ranking the generated attribute recommendations based on a combined transaction score and loyalty score, and generating a user interface including the set of N recommendations. The attribute recommendations are generated by a variational inference model configured using a transaction matrix and a loyalty matrix. BRIEF DESCRIPTION OF THE DRAWINGS The features and advantages of the present invention will be more fully disclosed in, or rendered obvious by the following detailed description of the preferred embodiments, which are to be considered together with the accompanying drawings wherein like numbers refer to like parts and further wherein: FIG. 1 illustrates a block diagram of a computer system, in accordance with some embodiments. FIG. 2 illustrates a network environment configured to provide categorical recommendations to a user system based, at least in part, on a temporal loyalty, in accordance with some embodiments. FIG. 3 a flowchart illustrating a method of providing personalized recommendations in a network interface, in accordance with some embodiments. FIG. 4 is a process flow illustrating various steps of the method of providing personalized recommendations in a network interface illustrated in FIG. 3, in accordance with some embodiments. FIG. 5 illustrates a modified collaborative filtering model, in accordance with some embodiments. FIG. 6 illustrates an iterative probabilistic pipeline, in accordance with some embodiments. FIGS. 7A-7C illustrate embeddings corresponding to product families generated by a Metapath2Vec algorithm and a modified collaborative filtering model, in accordance with some embodiments. DETAILED DESCRIPTION The description of the preferred embodiments is intended to be read in connection with the accompanying drawings, which are to be consi