# US 2020/0005310 A1 · Machine learning engine for fraud detection during cross-location online transaction processing

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

- Number: US 2020/0005310 A1
- Document ID: US20200005310A1
- Status: Published application
- Kind: A1
- Application number: 16/024,382
- Filed: 2018-06-29
- Published: 2020-01-02
- Earliest claimed: 2018-06-29
- Assignees: Paypal, Inc.
- Applicants: Paypal, Inc.
- Inventors: Dinesh Kumar; Geethamanjusha Melanathurbhaskar
- Web page: https://patentgrep.com/patent/US20200005310A1

## CPC

- G06Q 20/4016: {involving fraud or risk level assessment in transaction processing}
- G06F 18/23213: with fixed number of clusters, e.g. K-means clustering
- G06N 5/025: Extracting rules from data
- G06N 20/00: Machine learning
- G06Q 20/405: {Establishing or using transaction specific rules}
- G06Q 40/00: Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- 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)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

A machine learning engine for fraud detection related to cross-location online transaction processing may be trained using artificial intelligence techniques and used according to techniques discussed herein. An account may be used to electronically process a transaction for an item in a foreign location, such as a new city or country. The transaction may be identified as potentially fraudulent based on the item and/or location of purchase. A service provider may identify a vertical, such as an item type, for the transaction, and may determine the account's propensity to purchase within that vertical in the new location and the merchant's propensity to sell within that vertical to the account's location or shipping address. Based on the propensities, the service provider may utilize one or more risk rules with a risk assessment engine to determine transaction processing risk and whether to proceed with transaction processing.

## Claims

### Claim 1

A machine learning system for risk rule generation based on mined data association comprising:
a non-transitory memory storing a database of risk rules generated by a machine learning engine using past transactional data and an association rule mining process utilized by the machine learning engine; and
one or more hardware processors coupled to the non-transitory memory and configured to execute stored instructions to cause the machine learning system to perform operations comprising:
in response to receiving, from an online account, a processing request for a transaction with an entity located at a first location, determining a first vertical categorization for the transaction;
determining a second location associated with the online account based on the processing request;
identifying a vertical cluster associated with the online account based on the first location and the second location, wherein the vertical cluster comprises a second vertical categorization having a previous transaction completed by the online account within the second vertical categorization;
determining, using the machine learning engine, a vertical propensity for the online account to request the transaction within the first vertical categorization based on the second vertical categorization and one or more of the risk rules; and
generating, using the machine learning engine, a risk assessment of the transaction based on the vertical propensity.

### Claim 2 (depends on 1)

The machine learning system of claim 1, wherein the vertical propensity comprises a probability associated with purchase transactions by the online account and a plurality of other accounts associated with the second location for purchase of items within the first vertical categorization from entities at the first location.

### Claim 3 (depends on 2)

The machine learning system of claim 2, wherein prior to identifying the vertical cluster, the operations further comprise:
identifying shared verticals for the online account and the plurality of other accounts using at least the second vertical categorization; and
generating the vertical cluster comprising the online account and the plurality of other accounts based on the identifying.

### Claim 4 (depends on 3)

The machine learning system of claim 3, wherein the vertical cluster is generated based on clustering account transaction data for the online account and the plurality of other accounts using k-means clustering.

### Claim 5 (depends on 1)

The machine learning system of claim 1, wherein the operations further comprise:
determining a location rank for the first vertical categorization at the first location based on a plurality of transactions by the online account and a plurality of other accounts,
wherein the vertical propensity is further determined based on the location rank.

### Claim 6 (depends on 1)

The machine learning system of claim 1, wherein the operations further comprise:
determining that the transaction requires additional authentication to process the transaction with the online account based on the risk assessment; and
outputting an authentication request interface on a device associated with the processing request based on the additional authentication.

### Claim 7 (depends on 6)

The machine learning system of claim 6, wherein the operations further comprise:
in response to authenticating the device through the additional authentication, processing the transaction using the online account; and
updating the vertical propensity based on the processing the transaction.

### Claim 8 (depends on 1)

The machine learning system of claim 1, wherein the first location comprises a first country where the entity sells items within the first vertical categorization, wherein the second location comprises a second country of use of the online account, and wherein the transaction comprises a cross-border transaction with the entity using the online account.

### Claim 9 (depends on 1)

The machine learning system of claim 1, wherein the operations further comprise:
determining that the transaction is fraudulent based on the risk assessment; and
locking the online account from use based on the determining that the transaction is fraudulent.

### Claim 10 (depends on 1)

The machine learning system of claim 1, wherein the operations further comprise:
determining an account value level for the online account,
wherein the risk assessment is further based on the account value level.

### Claim 11 (depends on 1)

The machine learning system of claim 1, wherein the operations further comprise:
in response to performing the risk assessment, processing the transaction in accordance with the risk assessment; and
outputting a transaction processing result to a device associated with the processing request.

### Claim 12

A method comprising:
accessing a first user vertical for a first transaction type, wherein the first user vertical comprises a first number of first user transactions for the first transaction type from a first location by a user located at a second location;
determining a second user vertical for the user based on a vertical cluster for the user, wherein the second user vertical comprises a second number of second user transactions for a second transaction type from the first location by the user located at the second location;
determining a vertical propensity for the user to purchase an item associated with the first transaction type based on the first user vertical and the second user vertical; and
generating a fraud assessment rule based on the vertical propensity.

### Claim 13 (depends on 12)

The method of claim 12, further comprising:
accessing a merchant vertical for the first transaction type, wherein the merchant vertical comprises a third number of merchant sale transactions for the first transaction type by a merchant at the first location,
wherein the vertical propensity is further based on the merchant vertical.

### Claim 14 (depends on 12)

The method of claim 12, further comprising:
receiving a transaction request by the user for an item associated with the first transaction type from a merchant;
determining, by a fraud assessment engine, a fraud assessment for the transaction request based on the fraud assessment rule; and
processing the transaction request based on the fraud assessment, wherein the processing the transaction request comprises one of approving the transaction request when the fraud assessment is at or below a threshold risk level or denying the transaction request when the fraud assessment is above the threshold risk level.

### Claim 15 (depends on 12)

The method of claim 12, wherein the determining the vertical propensity uses association rule mining machine learning to identify an association between the first user vertical and the second user vertical, and wherein the fraud assessment rule comprises a probability for the user to purchase an item associated with the first user vertical from the first location while the user is located at the second location.

### Claim 16 (depends on 12)

The method of claim 12, wherein the first user transactions and the second user transactions comprise cross-border transactions from separate countries, and wherein the vertical propensity comprises a likelihood of a purchase of an item associated with the first user vertical based on the cross-border transactions.

### Claim 17

A non-transitory machine-readable medium having stored thereon machine-readable instructions executable to cause a machine to perform operations comprising:
accessing consumer transaction data for consumers, wherein the consumer transaction data comprises purchase transactions for item verticals at a plurality of locations by the consumers;
determining consumer vertical clusters based on the consumer transaction data, wherein each of the consumer vertical clusters comprises a vertical type and a first number of the purchase transactions in the vertical type;
generating, by a risk detection engine, a plurality of risk rules based on the consumer vertical clusters using association rule mining machine learning, wherein the plurality of risk rules comprises a risk rating for each of the consumer vertical clusters based on the first number of the purchase transactions at the plurality of locations; and
providing, by the risk detection engine, a transaction processing recommendation for a transaction based on the plurality of risk rules.

### Claim 18 (depends on 17)

The non-transitory machine-readable medium of claim 17, wherein the operations further comprise:
accessing merchant transaction data for merchants, wherein the consumer transaction data comprises sale transactions for the item verticals at the plurality of locations; and
determining merchant vertical clusters based on the merchant transaction data, wherein each of the merchant vertical clusters comprises the vertical type and a second number of the sale transactions in the vertical type,
wherein the plurality of risk rules are further generated using the merchant vertical clusters.

### Claim 19 (depends on 18)

The non-transitory machine-readable medium of claim 18, wherein each of the plurality of risk rules comprises a propensity for a purchase of an item in each of the item verticals by one of the consumers at a first location of the plurality of locations from one of the merchants at a second location of the plurality of locations.

### Claim 20 (depends on 19)

The non-transitory machine-readable medium of claim 19, wherein the operations further comprise:
receiving at least one of transaction processing feedback associated with a transaction, additional consumer transaction data, or additional merchant transaction data; and
updating, by the risk detection engine, the plurality of risk rules based on the at least one of the transaction processing feedback, the additional consumer transaction data, or the additional merchant transaction data.
