# US 2021/0174247 A1 · Calculating decision score thresholds using linear programming

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

- Number: US 2021/0174247 A1
- Document ID: US20210174247A1
- Status: Published application
- Kind: A1
- Application number: 16/709,504
- Filed: 2019-12-10
- Published: 2021-06-10
- Earliest claimed: 2019-12-10
- Assignees: Paypal, Inc.
- Applicants: PayPal, Inc.
- Inventors: Chiara Poletti; Hanlin Wu; Xing Ji; Moein Saleh
- Web page: https://patentgrep.com/patent/US20210174247A1

## CPC

- G06N 20/00: Machine learning
- G06F 16/22: Indexing; Data structures therefor; Storage structures
- G06Q 20/4016: {involving fraud or risk level assessment in transaction processing}
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- 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

## Abstract

A computer system is configured to receive a dataset that includes a plurality of transaction request records and is divisible into a plurality of segments. Each transaction request record includes an individual score calculated by a machine learning algorithm. The computer system also receives a plurality of constraints. The computer system is configured to calculate, using a linear programming algorithm, a decision threshold score for a particular segment of the plurality of segments using the transaction request records. The computer system is configured to provides access to the calculated decision threshold score to a production computer system. The production computer system is configured to use the decision threshold score to evaluate a subsequent transaction request corresponding to the particular segment.

## Claims

### Claim 1

A method comprising:
receiving, at a computer system, a dataset including a plurality of transaction request records, wherein the dataset is divisible into a plurality of segments of the dataset and wherein each transaction request record includes an individual score calculated by a machine learning algorithm;
receiving, at the computer system, a plurality of constraints;
calculating, with the computer system, a decision threshold score for a particular segment of the plurality of segments using the transaction request records, wherein the calculating includes using a linear integer programming algorithm to identify a decision threshold score for the particular segment according to the plurality of constraints; and
providing, by the computer system, access to the decision threshold score to a production computer system, wherein the decision threshold score is usable by the production computer system to evaluate a subsequent transaction request corresponding to the particular segment.

### Claim 2 (depends on 1)

The method of claim 1, wherein using a linear integer programming algorithm to calculate a decision threshold score includes:
projecting a plurality of performance indicators for transaction request records corresponding to the particular segment for each of a plurality of candidate decision threshold scores that satisfy the constraints; and
identifying, as the decision threshold score for the particular segment, the maximum candidate decision threshold score that satisfy the constraints.

### Claim 3 (depends on 1)

The method of claim 1, wherein the individual scores are fraud risk scores and the decision threshold score is a minimum fraud risk score threshold.

### Claim 4 (depends on 1)

The method of claim 1, wherein the plurality of constraints includes:
a transaction decline rate;
a catch rate; and
a false positive rate.

### Claim 5 (depends on 1)

The method of claim 1, wherein the plurality of constraints includes a global decline rate for the dataset and a segment decline rate for the particular segment.

### Claim 6 (depends on 1)

The method of claim 1, wherein each transaction request record is associated with a respective account and the dataset is divisible by levels of account activity.

### Claim 7 (depends on 1)

The method of claim 1, wherein calculating the decision threshold score for the particular segment includes:
calculating results for each of a plurality of candidate decision thresholds for each constraint; and
using linear integer programming to select an optimal candidate decision threshold for which the calculated results satisfy the constraints.

### Claim 8 (depends on 7)

The method of claim 7, wherein the constraints include a transaction decline rate and a catch rate, and wherein calculating results for each of a plurality of candidate decision thresholds includes calculating a transaction decline result and a catch result for each candidate decision threshold.

### Claim 9

A non-transitory, computer-readable medium storing instructions that when executed by a computer system cause the computer system to perform operations comprising:
receiving, at the computer system, a dataset including a plurality of transaction request records, wherein the dataset is divisible into a plurality of segments of the dataset and wherein each transaction request record includes an indication of whether the transaction request was granted;
receiving, at the computer system, a plurality of constraints;
calculating, with the computer system, a decision threshold score for a particular segment of the plurality of segments using the transaction request records, wherein the calculating includes using a linear integer programming algorithm to identify a decision threshold score for the particular segment according to the plurality of constraints; and
providing, by the computer system, access to the decision threshold score to a production computer system, wherein the decision threshold score is usable by the production computer system to evaluate a subsequent transaction request corresponding to the particular segment.

### Claim 10 (depends on 9)

The computer-readable medium of claim 9, wherein using a linear integer programming algorithm to calculate a decision threshold score includes:
projecting a plurality of performance indicators for transaction request records corresponding to the particular segment for each of a plurality of candidate decision threshold scores that satisfy the constraints; and
identifying, as the decision threshold score for the particular segment, the maximum candidate decision threshold score that satisfy the constraints.

### Claim 11 (depends on 9)

The computer-readable medium of claim 9, wherein each transaction request record includes individual scores that are fraud risk scores and wherein the decision threshold score is a minimum fraud risk score threshold.

### Claim 12 (depends on 9)

The computer-readable medium of claim 9, wherein the plurality of constraints includes:
a transaction decline rate;
a catch rate; and
a false positive rate.

### Claim 13 (depends on 9)

The computer-readable medium of claim 9, wherein the plurality of constraints includes a global catch rate for the dataset and a segment catch rate for the particular segment.

### Claim 14 (depends on 9)

The computer-readable medium of claim 9, wherein each transaction request record is associated with a respective geographic area and the dataset is divisible by geographic areas.

### Claim 15 (depends on 9)

The computer-readable medium of claim 9, wherein calculating the decision threshold score for the particular segment includes:
calculating results for each of a plurality of candidate decision thresholds for each constraint; and
using linear integer programming to select an optimal candidate decision threshold for which the calculated results satisfy the constraints.

### Claim 16

A method comprising:
running, on a production computer system, a transaction decision module operable to automatically determine whether to grant transaction requests based on one or more decision thresholds scores;
receiving, at the production computer system from a computer system, a particular decision threshold score corresponding to a particular type of transaction request, wherein the particular decision threshold score was calculated by:
analyzing a dataset including a plurality of transaction request records, wherein the dataset is divisible into a plurality of segments of the dataset and wherein each transaction request record includes an individual score calculated by a machine learning algorithm; and
calculating, with the computer system, the particular decision threshold for a particular segment of the plurality of segments, wherein the calculating includes using a linear integer programming algorithm to identify a decision threshold score for the particular segment according to a plurality of constraints;
receiving, at the production computer system, a particular transaction request of the particular type;
calculating, with the production computer system, a score for the particular transaction request using a machine learning algorithm; and
determining, with the production computer system, whether to grant the particular transaction request based on the score for the particular transaction request and the decision threshold score.

### Claim 17 (depends on 16)

The method of claim 16,
wherein the machine learning algorithm is trained prior to receiving particular transaction request; and
wherein the machine learning algorithm is not retrained in response to determining whether to grant the particular transaction request.

### Claim 18 (depends on 16)

The method of claim 16, wherein using a linear integer programming algorithm to calculate a decision threshold score includes:
projecting a plurality of performance indicators for transaction request records corresponding to the particular segment for each of a plurality of candidate decision threshold scores that satisfy the constraints; and
identifying, as the decision threshold score for the particular segment, the maximum candidate decision threshold score that satisfy the constraints.

### Claim 19 (depends on 16)

The method of claim 16, wherein the score for the particular transaction request is a fraud risk score and the decision threshold is a minimum fraud risk score threshold.

### Claim 20 (depends on 16)

The method of claim 16, wherein the plurality of constraints includes:
a transaction decline rate;
a catch rate; and
a false positive rate.
