Automatically generating similar items using spectral filtering

Patent metadata

Field	Value
Publication number	US-2022261873-A1
Application number	US-202117163510-A
Country	US
Kind code	A1
Filing date	Jan 31, 2021
Priority date	Jan 31, 2021
Publication date	Aug 18, 2022
Grant date	—

Abstract

A system including one or more processors and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform generating one or more item relational graphs for one or more items based on historical user purchases; transforming, using spectral filtering, the one or more item relational graphs into one or more frequency signals to remove noise from the one or more frequency signals; constructing, using a machine learning model, one or more item pair label classifications for one or more item pairs of the one or more items; generating a respective similarity score for each of the one or more item pairs; outputting a top k results for the one or more item pairs ranked by the respective similarity scores; and re-ranking, using a re-ranking algorithm, the top k results of the one or more item pairs based on a user preference for display on a user interface of an electronic device of a user. Other embodiments are disclosed.

First claim

What is claimed is:

1 . A system comprising:
one or more processors; and one or more non-transitory computer-readable media storing computing instructions configured to run on the one or more processors and perform:
generating one or more item relational graphs for one or more items based on historical user purchases;
transforming, using spectral filtering, the one or more item relational graphs into one or more frequency signals to remove noise from the one or more frequency signals;
constructing, using a machine learning model, one or more item pair label classifications for one or more item pairs of the one or more items;
generating a respective similarity score for each of the one or more item pairs;
outputting a top k results for the one or more item pairs ranked by the respective similarity scores; and
re-ranking, using a re-ranking algorithm, the top k results of the one or more item pairs based on a user preference for display on a user interface of an electronic device of a user.

2 . The system of claim 1 , wherein the computing instructions are further configured to run on the one or more processors and perform:
receiving, from users, a selection of a first item of the one or more items, wherein the first item is recommended to the users; generating a first node for the first item on an item relational graph of the one or more item relational graphs; and generating a second node for a recommendation for the first item, wherein the second node is connected to the first node on the item relational graph of the one or more item relational graphs, wherein an edge connecting the first node and the second node comprises a recommendation edge weight based on a number of times the recommendation for the first item is selected by the users over a period of time.

3 . The system of claim 1 , wherein the computing instructions are further configured to run on the one or more processors and perform:
receiving, from users, a selection of a third item of the one or more items, wherein the third item is substituted for a second item of the one or more items; generating a third node for the third item on an item relational graph of the one or more item relational graphs; and generating a fourth node for a substitution of the third item for the second item, wherein the fourth node is connected to the third node on the item relational graph of the one or more item relational graphs, wherein an edge connecting the third node and the fourth node comprises a substitution edge weight based on a number of times the substitution of the third item for the second item is selected by the users over a period of time.

4 . The system of claim 1 , wherein using the spectral filtering further comprises:
receiving an item relational graph of the one or more item relational graphs comprising two or more nodes connected by one or more edge weights, wherein the one or more edge weights represent one or more degrees of item-pair similarity.

5 . The system of claim 1 , wherein using the machine learning model further comprises:
training a data set based on input comprising historical item data over a period of time by:
encoding, using contextual feature encoding, the one or items.

6 . The system of claim 5 , wherein using contextual feature encoding further comprises:
using Poincare encoding.

7 . The system of claim 5 , wherein training the data set further comprises:
measuring a similarity, using a Jaccard similarity coefficient, of respective titles for each item within the one or more item pairs.

8 . The system of claim 1 , wherein the machine learning model is a feed-forward neural network.

9 . The system of claim 1 , wherein generating the respective similarity score further comprises:
outputting a probability that a recommendation or a substitute based on an item pair of the one or more item pairs will be selected by the users.

10 . The system of claim 1 , wherein:
using the re-ranking algorithm further comprises:
receiving affinity preferences for the user based on one or more features of an anchor item of the one or more item pairs in the top k results, as ranked;
constructing a respective affinity vector for each of the affinity preferences of the anchor item; and
scoring each item pair of the one or more item pairs, as ranked, based on the respective affinity vector of the anchor item; and
re-ranking the top k results further comprises:
re-ranking a subset of the top k results based on an affinity score for the user, wherein the user preference comprises the affinity score for the user.

11 . A method being implemented via execution of computing instructions configured to run on one or more processors and stored at one or more non-transitory computer-readable media, the method comprising:
generating one or more item relational graphs for one or more items based on historical user purchases; transforming, using spectral filtering, the one or more item relational graphs into one or more frequency signals to remove noise from the one or more frequency signals; constructing, using a machine learning model, one or more item pair label classifications for one or more item pairs of the one or more items; generating a respective similarity score for each of the one or more item pairs; outputting a top k results for the one or more item pairs ranked by the respective similarity scores; and re-ranking, using a re-ranking algorithm, the top k results of the one or more item pairs based on a user preference for display on a user interface of an electronic device of a user.

12 . The method of claim 11 , further comprising:
receiving, from users, a selection of a first item of the one or more items, wherein the first item is recommended to the users; generating a first node for the first item on an item relational graph of the one or more item relational graphs; and generating a second node for a recommendation for the first item, wherein the second node is connected to the first node on the item relational graph of the one or more item relational graphs, wherein an edge connecting the first node and the second node comprises a recommendation edge weight based on a number of times the recommendation for the first item is selected by the users over a period of time.

13 . The method of claim 11 , further comprising:
receiving, from users, a selection of a third item of the one or more items, wherein the third item is substituted for a second item of the one or more items; generating a third node for the third item on an item relational graph of the one or more item relational graphs; and generating a fourth node for a substitution of the third item for the second item, wherein the fourth node is connected to the third node on the item relational graph of the one or more item relational graphs, wherein an edge connecting the third node and the fourth node comprises a substitution edge weight based on a number of times the substitution of the third item for the second item is selected by the users over a period of time.

14 . The method of claim 11 , wherein using the spectral filtering further comprises:
receiving an item relational graph of the one or more item relational graphs comprising two or more nodes connected by one or more edge weights, wherein the one or more edge weights represent one or more degrees of item-pair similarity.

15 . The method of claim 11 , wherein using the machine learning model further comprises:
training a data set based on input comprising historical item data over a period of time by:
encoding, using contextual feature encoding,

Assignees

• Walmart Apollo Llc

Inventors

• Xu Da
• Mani Venugopal
• Ruan Chuanwei
• Kumar Sushant
• Achan Kannan

Classifications

• G06Q30/0201

  Market modelling; Market analysis; Collecting market data · CPC title

• G06Q30/0631 Primary

  Recommending goods or services · CPC title

• G06F16/212

  with details for data modelling support · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06N20/00

  Machine learning · CPC title