# US 2022/0036429 A1 · Systems and methods for generating recommendations using a corpus of data

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0036429 A1
- Document ID: US20220036429A1
- Status: Published application
- Kind: A1
- Application number: 17/502,839
- Filed: 2021-10-15
- Published: 2022-02-03
- Earliest claimed: 2015-12-11
- Assignees: Mastercard International Incorporated
- Applicants: Mastercard International Incorporated
- Inventors: Christopher John Merz; Walter F. Lo Faro; Peng Yang
- Web page: https://patentgrep.com/patent/US20220036429A1

## CPC

- G06Q 30/0631: {Recommending goods or services}
- G06F 7/08: Sorting, i.e. grouping record carriers in numerical or other ordered sequence according to the classification of at least some of the information they carry (by merging two or more sets of carriers in ordered sequence G06F7/16)
- G06Q 30/0255: {based on user history}
- G06Q 30/0282: Rating or review of business operators or products
- G06Q 30/0625: {by formulating product or service queries, e.g. using keywords or predefined options}
- G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method and system for recommending a merchant are provided. The method includes receiving financial transaction data documenting financial transactions between a plurality of account holders and a plurality of merchants and generating a merchant correspondence matrix that includes the plurality of merchants and a plurality of indicators of interactions associated with pairs of the plurality of merchants. The plurality of indicators of interactions tallying financial transactions conducted by the plurality of account holders at both of the merchants in a pair of the plurality of merchants. The method further includes receiving a query for a recommendation of a merchant from an account holder and generating a ranked list of merchants based on a recommender algorithm. The recommender algorithm inferring user preferences from attributes of the plurality of merchants that were visited by the cardholder.

## Claims

### Claim at position 1

1-20. (canceled)

### Claim 21

A computer-based method for predicting preferences, the method implemented using a recommender computer device including at least one processor coupled to a memory device, the method comprising steps performed by the at least one processor comprising:
accessing a database storing communications routed by a payment card network between a plurality of acquirer computing devices and a plurality of issuer computing devices, each of the communications formatted according to a standard for exchange of financial transaction data and settlement of funds between financial institutions that are customers of the payment card network, each of the communications corresponding to a transaction between a respective one of a plurality of merchants and a respective one of a plurality of consumers, each of the communications including data identifying the respective merchant and an account of the respective consumer;
generating, in the memory device for a list of merchants included in the plurality of merchants, an array of values (i,j) corresponding to a merchant correspondence matrix, wherein the list comprises an ordered sequence of merchants numbered 1 to n, i and j are indices, the index i ranges from 1 to n, and the index j ranges from 1 to n, said generating comprising:
searching the data identifying the respective merchant in the communications to identify transactions associated with the merchants in the list of merchants;
identifying, from the communications associated with the identified transactions, the accounts that transacted at each merchant in the list of merchants;
populating, in the memory device, each of the values (i,j) of the array based on a number of co-visits of merchant i and merchant j, wherein each co-visit is defined as a respective one of the identified accounts having at least one of the identified transactions at merchant i and at least one of the identified transactions at merchant j;
multiplying the merchant correspondence matrix and a uniform preference vector to obtain a first results vector, the uniform preference vector having n equally weighted elements;
determining an initial rank of each merchant in the list based on the values in the first results vector,
receiving, from a client computing device via the Internet, a registration from a first consumer of the plurality of consumers, the first consumer associated with a first account;
searching the data identifying the respective account in the communications to identify transactions associated with the first account;
identifying, from the communications associated with the first account, each merchant in the list of merchants at which the first account transacted;
generating a user preference vector for the first account, the user preference vector comprising n elements, wherein elements corresponding in sequence to one of the visited merchants are weighted heavier than the elements of the user preference vector that do not correspond to one of the visited merchants;
receiving, from the client computing device via the Internet, a query from the first consumer for a recommendation of a merchant;
generating a second results vector, comprising multiplying, by the at least one processor, the merchant correspondence matrix and the user preference vector;
determining an updated rank of each merchant in the list based on the values in the second results vector;
predicting a ranked preference of the first consumer for the merchants in the list of merchants based on a change in rank of merchants in the list between the initial rank and the updated rank; and
identifying, to the client computing device via the Internet, one or more merchants based on the predicted ranked preference.

### Claim 22 (depends on 21)

The computer-based method of claim 21, wherein said multiplying the merchant correspondence matrix and the user preference vector generates an intermediate input vector, and said generating the second results vector further comprises:
a) obtaining an iteration result vector comprising multiplying the merchant correspondence matrix and the intermediate input vector;
b) updating the intermediate input vector to equal the iteration result vector;
repeating steps a and b until an end condition is satisfied; and
in response to the end condition being satisfied, setting the second results vector equal to the iteration result vector.

### Claim 23 (depends on 22)

The computer-based method of claim 22, wherein the end condition is a completion of at least five iterations.

### Claim 24 (depends on 23)

The computer-based method of claim 23, wherein said generating the second results vector comprises pre-computing the second results vector prior to said receiving the query.

### Claim 25 (depends on 21)

The computer-based method of claim 21, wherein said method is performed for the list having n greater than or equal to 30,000.

### Claim 26 (depends on 21)

The computer-based method of claim 21, wherein said predicting the ranked preference based on the change in rank comprises assigning each merchant's preference ranking based on a number of positions the merchant has changed between the initial rank and the updated rank.

### Claim 27 (depends on 21)

The computer-based method of claim 21, wherein said predicting the ranked preference of the first consumer comprises predicting the ranked preference of the first consumer for the list of merchants in a first metropolitan area, said method further comprising:
using a portion of top-ranked merchants in the predicted rank preference to seed a second merchant correspondence matrix; and
predicting a ranked preference of the first consumer for a second list of merchants in a second metropolitan area using the seeded second merchant correspondence matrix.

### Claim 28

A preference predictor system comprising at least one processor coupled to a memory device, said memory device including instructions executable to cause the at least one processor to perform steps including:
accessing a database storing communications routed by a payment card network between a plurality of acquirer computing devices and a plurality of issuer computing devices, each of the communications formatted according to a standard for exchange of financial transaction data and settlement of funds between financial institutions that are customers of the payment card network, each of the communications corresponding to a transaction between a respective one of a plurality of merchants and a respective one of a plurality of consumers, each of the communications including data identifying the respective merchant and an account of the respective consumer;
generating, in the memory device for a list of merchants included in the plurality of merchants, an array of values (i,j) corresponding to a merchant correspondence matrix, wherein the list comprises an ordered sequence of merchants numbered 1 to n, i and j are indices, the index i ranges from 1 to n, and the index j ranges from 1 to n, the generating comprising:
searching the data identifying the respective merchant in the communications to identify transactions associated with the merchants in the list of merchants;
identifying, from the communications associated with the identified transactions, the accounts that transacted at each merchant in the list of merchants;
populating, in the memory device, each of the values (i,j) of the array based on a number of co-visits of merchant i and merchant j, wherein each co-visit is defined as a respective one of the identified accounts having at least one of the identified transactions at merchant i and at least one of the identified transactions at merchant j;
multiplying the merchant correspondence matrix and a uniform preference vector to obtain a first results vector, the uniform preference vector having n equally weighted elements;
determining an initial rank of each merchant in the list based on the values in the first results vector,
receiving, from a client computing device via the Internet, a registration from a first consumer of the plurality of consumers, the first consumer associated with a first account;
searching the data identifying the respective account in the communications to identify transactions associated with the first account;
identifying, from the communications associated with the first account, each merchant in the list of merchants at which the first account transacted;
generating a user preference vector for the first account, the user preference vector comprising n elements, wherein elements corresponding in sequence to one of the visited merchants are weighted heavier than the elements of the user preference vector that do not correspond to one of the visited merchants;
receiving, from the client computing device via the Internet, a query from the first consumer for a recommendation of a merchant;
generating a second results vector, comprising multiplying, by the at least one processor, the merchant correspondence matrix and the user preference vector;
determining an updated rank of each merchant in the list based on the values in the second results vector;
predicting a ranked preference of the first consumer for the merchants in the list of merchants based on a change in rank of merchants in the list between the initial rank and the updated rank; and
identifying, to the client computing device via the Internet, one or more merchants based on the predicted ranked preference.

### Claim 29 (depends on 28)

The preference predictor system of claim 28, wherein the instructions are further executable to cause the at least one processor to:
multiply the merchant correspondence matrix and the user preference vector to generate an intermediate input vector; and
generate the second results vector by performing further steps including:
a) obtaining an iteration result vector comprising multiplying the merchant correspondence matrix and the intermediate input vector;
b) updating the intermediate input vector to equal the iteration result vector;
repeating steps a and b until an end condition is satisfied; and
in response to the end condition being satisfied, setting the second results vector equal to the iteration result vector.

### Claim 30 (depends on 29)

The preference predictor system of claim 29, wherein the end condition is a completion of at least five iterations.

### Claim 31 (depends on 30)

The preference predictor system of claim 30, wherein the instructions are further executable to cause the at least one processor to pre-compute the second results vector prior to receiving the query.

### Claim 32 (depends on 28)

The preference predictor system of claim 28, wherein n is greater than or equal to 30,000.

### Claim 33 (depends on 28)

The preference predictor system of claim 28, wherein the instructions are further executable to cause the at least one processor to predict the ranked preference based on the change in rank by additional steps including assigning each merchant's preference ranking based on a number of positions the merchant has changed between the initial rank and the updated rank.

### Claim 34 (depends on 28)

The preference predictor system of claim 28, wherein the step of predicting the ranked preference of the first consumer comprises predicting the ranked preference of the first consumer for the list of merchants in a first metropolitan area, and the instructions are further executable to cause the at least one processor to perform additional steps including:
using a portion of top-ranked merchants in the predicted rank preference to seed a second merchant correspondence matrix; and
predicting a ranked preference of the first consumer for a second list of merchants in a second metropolitan area using the seeded second merchant correspondence matrix.

### Claim 35

One or more non-transitory computer-readable storage media having computer-executable instructions embodied thereon for execution by at least one processor of a preference predictor system, the at least one processor in communication with a memory device, the computer-executable instructions being executable to cause the at least one processor to perform steps including:
accessing a database storing communications routed by a payment card network between a plurality of acquirer computing devices and a plurality of issuer computing devices, each of the communications formatted according to a standard for exchange of financial transaction data and settlement of funds between financial institutions that are customers of the payment card network, each of the communications corresponding to a transaction between a respective one of a plurality of merchants and a respective one of a plurality of consumers, each of the communications including data identifying the respective merchant and an account of the respective consumer;
generating, in the memory device for a list of merchants included in the plurality of merchants, an array of values (i,j) corresponding to a merchant correspondence matrix, wherein the list comprises an ordered sequence of merchants numbered 1 to n, i and j are indices, the index i ranges from 1 to n, and the index j ranges from 1 to n, said generating comprising:
searching the data identifying the respective merchant in the communications to identify transactions associated with the merchants in the list of merchants;
identifying, from the communications associated with the identified transactions, the accounts that transacted at each merchant in the list of merchants;
populating, in the memory device, each of the values (i,j) of the array based on a number of co-visits of merchant i and merchant j, wherein each co-visit is defined as a respective one of the identified accounts having at least one of the identified transactions at merchant i and at least one of the identified transactions at merchant j;
multiplying the merchant correspondence matrix and a uniform preference vector to obtain a first results vector, the uniform preference vector having n equally weighted elements;
determining an initial rank of each merchant in the list based on the values in the first results vector,
receiving, from a client computing device via the Internet, a registration from a first consumer of the plurality of consumers, the first consumer associated with a first account;
searching the data identifying the respective account in the communications to identify transactions associated with the first account;
identifying, from the communications associated with the first account, each merchant in the list of merchants at which the first account transacted;
generating a user preference vector for the first account, the user preference vector comprising n elements, wherein elements corresponding in sequence to one of the visited merchants are weighted heavier than the elements of the user preference vector that do not correspond to one of the visited merchants;
receiving, from the client computing device via the Internet, a query from the first consumer for a recommendation of a merchant;
generating a second results vector, comprising multiplying, by the at least one processor, the merchant correspondence matrix and the user preference vector;
determining an updated rank of each merchant in the list based on the values in the second results vector;
predicting a ranked preference of the first consumer for the merchants in the list of merchants based on a change in rank of merchants in the list between the initial rank and the updated rank; and
identifying, to the client computing device via the Internet, one or more merchants based on the predicted ranked preference.

### Claim 36 (depends on 35)

The one or more non-transitory computer-readable storage media of claim 35, wherein the instructions are further executable to cause the at least one processor to:
multiply the merchant correspondence matrix and the user preference vector to generate an intermediate input vector; and
generate the second results vector by performing further steps including:
a) obtaining an iteration result vector comprising multiplying the merchant correspondence matrix and the intermediate input vector;
b) updating the intermediate input vector to equal the iteration result vector;
repeating steps a and b until an end condition is satisfied; and
in response to the end condition being satisfied, setting the second results vector equal to the iteration result vector.

### Claim 37 (depends on 36)

The one or more non-transitory computer-readable storage media of claim 36, wherein the end condition is a completion of at least five iterations.

### Claim 38 (depends on 37)

The one or more non-transitory computer-readable storage media of claim 37, wherein the instructions are further executable to cause the at least one processor to pre-compute the second results vector prior to receiving the query.

### Claim 39 (depends on 35)

The one or more non-transitory computer-readable storage media of claim 35, wherein n is greater than or equal to 30,000.

### Claim 40 (depends on 35)

The one or more non-transitory computer-readable storage media of claim 35, wherein the step of predicting the ranked preference of the first consumer comprises predicting the ranked preference of the first consumer for the list of merchants in a first metropolitan area, and the instructions are further executable to cause the at least one processor to perform additional steps including:
using a portion of top-ranked merchants in the predicted rank preference to seed a second merchant correspondence matrix; and
predicting a ranked preference of the first consumer for a second list of merchants in a second metropolitan area using the seeded second merchant correspondence matrix.
