# US 2022/0027984 A1 · Systems and methods for automating pricing desk operation

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

- Number: US 2022/0027984 A1
- Document ID: US20220027984A1
- Status: Published application
- Kind: A1
- Application number: 17/443,159
- Filed: 2021-07-21
- Published: 2022-01-27
- Earliest claimed: 2020-07-22
- Assignees: Banque Nationale du Canada
- Applicants: Banque Nationale du Canada
- Inventors: Eric Charton; Yann-Edern L'hour; Marie-Eve Malette; Julien Rizkallah-Leroux; Michel Descoteaux
- Web page: https://patentgrep.com/patent/US20220027984A1

## CPC

- G06Q 40/02: Banking, e.g. interest calculation or account maintenance (credit or loans G06Q40/03)
- G06N 20/00: Machine learning
- G06N 20/20: Ensemble learning
- G06Q 30/0206: {Price or cost determination based on market factors}
- G06N 5/01: Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
- 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
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

According to an aspect, a system for automating pricing desk operations is described. The system includes an input module configured to receive a proposed offer in relation to a financial product and associated context information; a machine learning module trained to classify offers according to positive or negative client; a processing module configured to optimize the proposed and an output module configured to transmit the optimized offer for communication to the client. According to additional aspects, a corresponding computer-implemented method, non-transitory computer-readable medium, and systems and methods for training a machine learning module are also described.

## Claims

### Claim 1

A system for automating pricing desk operations, comprising:
an input module configured to receive a proposed offer to present to a client, and to receive input context information associated with the proposed offer, wherein the proposed offer comprises offer parameters in relation to a financial product;
a machine learning module trained using historical offer data to classify offers according to positive or negative client decisions, based on offer parameters and historical context information associated with the offers;
a processing module configured to optimize the proposed offer by performing the following subprocesses:
generating a modified offer by altering the offer parameters of the proposed offer;
processing the modified offer and the input context information using the machine learning module to determine an acceptance probability corresponding to a probability that the modified offer will result in a positive client decision; and
repeating the subprocesses of generating the modified offer and processing the modified offer to obtain an optimized offer having optimized offer parameters and an acceptance probability above a predetermined threshold; and
an output module configured to transmit the optimized offer for communication to the client.

### Claim 2 (depends on 1)

The system according to claim 1, wherein the processing module is configured to optimize the proposed offer by searching for an optimal offer from a generated set of modified offers, the subprocesses performed by the processing module comprising:
i) generating the set of modified offers, the set of modified offers corresponding to a range centered around the proposed offer;
ii) processing at least some of the modified offers in the set along with the input context information using the machine learning module to determine an acceptance probability corresponding to a probability that the modified offer will result in a positive client decision;
iii) identifying one of the modified offers having optimized offer parameters among the set of modified offers and an acceptance probability above the predetermined threshold, the identified one of the modified offers thereby corresponding to the optimized offer.

### Claim 3 (depends on 1)

The system according to claim 1, wherein the processing module is configured to optimize the proposed offer by iteratively modifying the proposed offer, the subprocesses performed by the processing module comprising:
i) assigning the proposed offer to a current offer;
ii) generating the modified offer by incrementally adjusting parameters of the current offer;
iii) processing the modified offer and the input context information using the machine learning module to obtain the acceptance probability of the modified offer;
iv) if the acceptance probability of the modified offer is above a predetermined threshold and the parameters of the modified offer are superior to the current offer, assigning the modified offer as the current offer and returning to subprocess ii); and
v) outputting the current offer as the optimized offer if the current offer cannot be further modified to have superior parameters while keeping the acceptance probability above the predetermined threshold.

### Claim 4 (depends on 1)

The system according to claim 1, wherein the offer parameters comprise an interest rate.

### Claim 5 (depends on 1)

The system according to claim 1, wherein the financial product is a mortgage.

### Claim 6 (depends on 1)

The system according to claim 1, wherein the historical context information and the input context information comprise client-specific information.

### Claim 7 (depends on 6)

The system according to claim 6, wherein the client-specific information comprises at least one of: financial transaction history, account balances, credit score, website interaction data, social media interaction data, previous applications for financial products, home address, work address, family and marital status, previous offers from competitors, and cost to break existing contract.

### Claim 8 (depends on 1)

The system according to claim 1, wherein the historical context information and the input context information comprise market-related information.

### Claim 9 (depends on 8)

The system according to claim 8, wherein the market-related information comprises at least one of: published competitor interest rates, daily costs of funds, and pricing targets.

### Claim 10 (depends on 1)

The system according to claim 1, wherein the historical context information and the input context information comprise product-related information, said product-related information corresponding to characteristics of a type of the financial product to which the offers relate.

### Claim 11 (depends on 1)

The system according to claim 1, wherein the machine learning module is trained using the historical offer data comprising feature vectors of the historical offer parameters and historical context information, each feature vector labeled as a positive or negative client decision.

### Claim 12 (depends on 1)

The system according to claim 1, wherein the machine learning module is retrained using the historical offer data at regular predetermined intervals.

### Claim 13 (depends on 1)

The system according to claim 1, wherein the machine learning module is trained using historical offer data from within a predetermined time period.

### Claim 14 (depends on 1)

The system according to claim 1, wherein the machine learning module is configured to process offers using a trained classification algorithm.

### Claim 15 (depends on 14)

The system according to claim 14, wherein the trained classification algorithm is selected from a group consisting of: logistic regression, random forest classifier, and gradient boosting classifier.

### Claim 16 (depends on 1)

The system according to claim 1, wherein the input module is configured to receive at least some of the historical context information associated or the input context information from at least one external database.

### Claim 17 (depends on 16)

The system according to claim 16, wherein the at least one external database comprises: a database of historical pricing desk offers, a database of historically accepted offer parameters, a database of client information, a database of historical rates offered by competitors, and a database of historical cost structures used to calculate offer parameters.

### Claim 18 (depends on 1)

The system according to claim 1, further comprising a preprocessing module configured to process the proposed offer and input context information before they are provided to the machine learning module.

### Claim 19 (depends on 18)

The system according to claim 18, wherein the preprocessing module is configured to perform at least one of: parsing and cleaning, joining additional information, formatting and validating, and scaling and encoding.

### Claim 20 (depends on 1)

The system according to claim 1, further comprising a request evaluation module configured to conduct preliminary subprocesses of:
receiving a request comprising requested parameters in relation to a financial product;
determining whether to authorize or refuse the request based on predetermined criteria; and
if the request is refused, initiating the generating of a counteroffer by providing the request or a modified request to the input module as a proposed offer for optimization.

### Claim 21 (depends on 20)

The system according to claim 20, wherein:
if the request is refused, the counteroffer is generated by generating an optimized offer having the acceptance probability above a first predetermined threshold; and
if the request is accepted, initiating the generating of the counteroffer by providing the request or a modified request to the input module as a proposed offer for optimization, the counteroffer being generated by generating an optimized offer having the acceptance probability above a second predetermined threshold that is greater than the first predetermined threshold.

### Claim 22 (depends on 1)

The system according to claim 1, further comprising a counteroffer generating module configured to conduct preliminary subprocesses of:
generating a counteroffer to a refused request from a client, and;
providing the counteroffer to the input module as a proposed offer for optimization.

### Claim 23 (depends on 22)

The system according to claim 22, wherein the counteroffer generating module is configured to generate the counteroffer using an artificial intelligence model.

### Claim 24

A computer-implemented method for automating pricing desk operations, comprising:
receiving, via an input module, a proposed offer to present to a client, wherein the proposed offer comprises offer parameters in relation to a financial product;
receiving, via the input module, input context information associated with the proposed offer;
optimizing the proposed offer using a processing module and a machine learning module trained using historical offer data to classify offers according to positive or negative client decisions, based on historical offer parameters and historical context information associated with the historical offers, wherein the proposed offer is optimized by performing the following subprocesses:
generating, via the processing module, a modified offer by altering the offer parameters of the proposed offer;
processing the modified offer and the input context information using the machine learning module to determine an acceptance probability corresponding to a probability that the modified offer will result in a positive client decision; and
repeating the subprocesses of generating the modified offer and processing the modified offer using the machine learning module to obtain an optimized offer having optimized offer parameters and an acceptance probability above a predetermined threshold; and
transmitting the optimized offer for communication to the client.

### Claim 25

A non-transitory computer-readable medium having instructions stored thereon which, when executed by a processor, cause the processor to:
receive a proposed offer to present to a client, wherein the proposed offer comprises offer parameters in relation to a financial product;
receive input context information associated with the proposed offer;
optimize the proposed offer using a machine learning module trained using historical offer data to classify offers according to positive or negative client decisions, based on historical offer parameters and historical context information associated with the historical offers, wherein the proposed offer is optimized by performing the following subprocesses:
generating a modified offer by altering the offer parameters of the proposed offer;
processing the modified offer and the input context information using the machine learning module to determine an acceptance probability corresponding to a probability that the modified offer will result in a positive client decision; and
repeating the subprocesses of generating the modified offer and processing the modified offer using the machine learning module to obtain an optimized offer having optimized offer parameters and an acceptance probability above a predetermined threshold; and
transmit the optimized offer for communication to the client.
