Systems and methods for using a predictive engine to predict failures in machine-learning trained systems for display via graphical user interface

Patent metadata

Field	Value
Publication number	US-12001927-B2
Application number	US-201916659960-A
Country	US
Kind code	B2
Filing date	Oct 22, 2019
Priority date	Oct 22, 2019
Publication date	Jun 4, 2024
Grant date	Jun 4, 2024

Abstract

A computer-implemented method for using a machine-learning trained predictive engine to predict failures, comprising receiving, from a user using a graphical user interface, electronic transaction data comprising an item and candidate transaction terms, the candidate transaction terms comprising a first transaction term, the electronic transaction data being associated with a candidate transaction. The method may determine, by a machine learning predictive engine, a likelihood of success of the candidate transaction based on the electronic transaction data, and may determine, by the machine learning predictive engine, based on the first transaction term, a second transaction term that, together with the first transaction term, increases the likelihood of success of the candidate transaction above a predetermined threshold. An indication of the second transaction term and an indication of the likelihood of success of the candidate transaction may be displayed on the graphical user interface.

First claim

What is claimed is:

1. A computer-implemented method for using a machine-learning predictive engine to predict failures in transaction occurrences, comprising:
receiving, from a user using a graphical user interface, electronic transaction data comprising an item and candidate transaction terms, the candidate transaction terms comprising a first transaction term, the electronic transaction data being associated with a candidate transaction involving a plurality of parties;
determining a first likelihood value indicative of a likelihood that all of the parties in the plurality of parties will collectively agree to the candidate transaction terms based on the electronic transaction data by comparing the candidate transaction terms using the machine learning predictive engine, wherein the machine learning predictive engine is trained based on electronic prior transaction data corresponding to terms of a plurality of prior transactions that occurred and terms for a plurality of prior candidate transactions that did not occur, such that the machine learning predictive engine is configured to determine the first likelihood value based on similarities between (i) the electronic transaction data and (ii) the plurality of prior transactions that occurred and the plurality of prior transactions that did not occur, respectively;
determining, by the machine learning predictive engine, based on the first transaction term, a second transaction term that, together with the first transaction term, increases the first likelihood value that all of the parties will agree to the candidate transaction terms to a second likelihood value that all of the parties will agree to the candidate transaction terms that is above a predetermined threshold; and
causing the graphical interface to display an indication of the first likelihood value, an indication of the second transaction term, and an indication of the second likelihood value via the graphical user interface.

2. The computer-implemented method of claim 1 , further comprising:
receiving the electronic prior transaction data corresponding to the plurality of prior transactions that occurred and the plurality of prior candidate transactions that did not occur; and
training the machine learning predictive engine based on the electronic prior transaction data corresponding to the plurality of prior transactions that occurred and the plurality of prior candidate transactions that did not occur.

3. The computer-implemented method of claim 2 , wherein the training the machine learning predictive engine further comprises determining a weight for each of the candidate transaction terms based on a contribution to the first likelihood value.

4. The computer-implemented method of claim 2 , wherein the electronic prior transaction data is received from a first data source, and wherein the training the machine learning predictive engine further comprises:
upon determining that the plurality of prior transactions that did occur or the plurality of prior transactions that did not occur is below a further predetermined threshold, automatically acquiring additional electronic prior transaction data from a second data source, the second data source being selected based upon similarity to the first data source.

5. The computer-implemented method of claim 1 , further comprising:
upon determining the second transaction term of the candidate transaction terms, determining a first effect of the first transaction term on the first likelihood value and a second effect of the second transaction term on the second likelihood value; and
causing the graphical user interface to display a first indicator of the first effect and a second indicator of the second effect.

6. The computer-implemented method of claim 5 , further comprising:
in response to receiving an update of the first transaction term from the graphical user interface, updating, by the machine learning predictive engine, the first effect on the first likelihood value; and
causing the graphical user interface to display the first indicator of the updated first effect.

7. The computer-implemented method of claim 1 , wherein the electronic transaction data comprises a plurality of historical candidate transaction terms proposed by the user, and further comprising:
determining, by the machine learning predictive engine, a range of acceptable candidate transaction terms based on values associated with the historical candidate transaction terms;
receiving additional electronic transaction data comprising additional candidate transaction terms;
determining, by the machine learning predictive engine, that values associated with the additional candidate transaction terms are outside of the range of acceptable candidate transaction terms by greater than a further predetermined threshold; and
causing the graphical interface to display an alert to the user that values associated with the additional candidate transaction terms are outside of the range of acceptable candidate transaction terms.

8. The computer-implemented method of claim 1 , further comprising:
determining whether the first likelihood value is below a first predetermined threshold;
in response to determining that the first likelihood value is below the first predetermined threshold, determining one or more alternate terms of the candidate transaction that increase the first likelihood value above the first predetermined threshold; and
causing the graphical interface to display one or more indicators of the one or more alternate terms of the candidate transaction on the graphical user interface.

9. The computer-implemented method of claim 1 , further comprising:
determining whether the first likelihood value is below a second predetermined threshold;
in response to determining that the first likelihood value is below the second predetermined threshold, determining one or more alternate items of the candidate transaction that increase the first likelihood value that all of the parties will agree to the candidate transaction terms above the second predetermined threshold; and
causing the graphical interface to display the one or more alternate items of the candidate transaction.

10. The computer-implemented method of claim 1 , further comprising:
determining a lower predetermined threshold;
determining a higher predetermined threshold, the higher predetermined threshold being higher than the lower predetermined threshold;
in response to determining that the first likelihood value is below the higher predetermined threshold and above the lower predetermined threshold, determining one or more alternate terms of the candidate transaction that increase the first likelihood value above the higher predetermined threshold;
in response to determining that the first likelihood value is below the higher predetermined threshold and below the lower predetermined threshold, determining one or more alternate items of the candidate transaction that increase the first likelihood value above the lower predetermined threshold; and
causing the graphical interface to display the one or more alternate terms or the one or more alternate items of the candidate transaction.

11. The computer-implemented method of claim 1 , further comprising:
receiving a selection, from the user, of one of a plurality of candidate term packages; and
based on the selection of the one of the plurality of candidate term packages, determining, by the machine learning predictive engine, a second plurality of candidate term packages for display to the user.

12. A computer-implemented method for using a machine-learning predictive engine to predict failures in transaction occurre

Assignees

• Capital One Services Llc

Inventors

• Hoover Jason
• Price Micah
• Wylie Stephen
• Tang Qiaochu
• Dagley Geoffrey

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06Q50/188 Primary

  Electronic negotiation · CPC title

• G06N5/046

  Forward inferencing; Production systems · CPC title