Automated systems for reducing computational loads in the mass execution of analytical models using scale-out computing

Patent metadata

Field	Value
Publication number	US-9978038-B2
Application number	US-201715709231-A
Country	US
Kind code	B2
Filing date	Sep 19, 2017
Priority date	Dec 18, 2015
Publication date	May 22, 2018
Grant date	May 22, 2018

Abstract

Disclosed embodiments provide systems and techniques for mass execution of analytical models across multiple dimensions of client, collateral, deal structure, third party, and other data relevant to predicting optimal decisions in real-time. In some embodiments, disclosed systems and techniques increase decisioning speed through the reduction of computational loads on disclosed decisioning systems. Further disclosed systems and techniques may scale-out analytical modeling computations through, among other technological solutions, advanced execution environments that are asynchronous and non-blocking in nature so as to allow the execution of a plurality of analytical models in parallel and optimizing the results.

First claim

What is claimed is:

1. A system for reducing computational loads associated with predicting transactions using scale-out computing of analytical models, comprising:
a memory storing instructions; and
one or more processors configured to execute the instructions to perform operations comprising:
receiving, via an Application Programming Interface (API), a transaction proposal including a first plurality of transaction parameters;
identifying, via an actor pool associated with a data interface module, a plurality of data sources, the plurality of data sources being identified based on the transaction proposal;
iteratively verifying, via an actor pool associated with the data interface module, the plurality of data sources;
generating, via a minimally viable transaction (MVT) generator, an MVT, the MVT comprising a second plurality of transaction parameters and being based on the transaction proposal;
generating, via an actor pool associated with a transaction generator, a first plurality of transaction options, the first plurality of transaction options being generated based on the transaction proposal and the MVT;
scoring, via an actor pool associated with a scoring engine, the MVT and at least one of the first plurality of transaction options, the MVT and the transaction options being scored using an analytical model;
verifying, via an actor pool associated with a policy calculator, that the MVT and the transaction options comply with a transaction policy, the MVT and the transaction options being verified based on the scoring;
optimizing, via an actor pool associated with a transaction optimizer, the transaction options based on a preference; and
transmitting, via the API:
data indicating whether the transaction proposal is approved, wherein whether the transaction proposal is approved is determined based on whether the MVT complies with the transaction policy, and
at least one transaction offer based on the optimized transaction options.

2. The system of claim 1 , wherein when the transaction proposal is rejected,
the transaction generator is further configured to cause the one or more processors to generate, via the actor pool associated with the transaction generator, a second plurality of transaction options, the second plurality of transaction options being generated based on the transaction proposal and the MVT;
the scoring engine is further configured to cause the one or more processors to score, via the actor pool associated with the scoring engine, at least one of the second plurality of transaction options, the at least one of the second plurality of transaction options being scored using the analytical model; and
the policy calculator is further configured to cause the one or more processors to verify, via the actor pool associated with the policy calculator, that at least one of the second plurality of transaction options complies with the transaction policy, the at least one of the second plurality of transaction options being verified based on the scoring,
wherein the API is further configured to cause the one or more processors to transmit the at least one of the second plurality of transaction options that complies with the transaction policy.

3. The system of claim 2 , wherein the API is further configured to cause the one or more processors to transmit a ranking associated with each of the at least one of the second plurality of transaction options.

4. The system of claim 1 , wherein at least one of the API, the data interface module, the MVT generator, the transaction generator, the scoring engine, the policy calculator, or the transaction optimizer comprise software modules.

5. The system of claim 4 , wherein the software modules are operated on a remote system.

6. The system of claim 1 , wherein the MVT represents a maximum extent for varying the first plurality of transaction parameters.

7. The system of claim 1 , wherein the MVT generator comprises instructions that cause the one or more processors to:
identify a variation increment for each of the first plurality of transaction parameters; and
determine a plurality of transaction parameter variations between the first plurality of transaction parameters and second plurality of transaction parameters according to the variation increment.

8. The system of claim 1 , wherein the analytical model comprises an applicant rating (AR) model and a structure rating (SR) model.

9. The system of claim 1 , wherein the data interface module, the transaction generator, the scoring engine, the policy calculator, and the transaction optimizer comprise fault-tolerant and scalable applications.

10. The system of claim 1 , wherein the data interface module, the transaction generator, the scoring engine, the policy calculator, and the transaction optimizer are implemented in at least one of a big data architecture, actor architecture, message bus architecture, or Lambda-based architecture.

11. The system of claim 1 , wherein at least a portion of the operations are performed in parallel.

12. The system of claim 1 , wherein at least one of the actor pools is configured to expand or contract based on a computational load associated with the operations associated with the at least one of the actor pools.

13. A method for reducing computational loads associated with predicting optimal transactions using scale-out computing of analytical models, comprising:
receiving, via an Application Programming Interface (API), a transaction proposal including a first plurality of transaction parameters;
identifying a plurality of data sources, the plurality of data sources being identified based on the transaction proposal;
iteratively verifying the plurality of data sources;
generating a minimally viable transaction (MVT), the MVT comprising a second plurality of transaction parameters and being based on the transaction proposal;
generating a first plurality of transaction options, the first plurality of transaction options being generated based on the transaction proposal and the MVT;
scoring the MVT and at least one of the first plurality of transaction options, the MVT and the transaction options being scored using an analytical model;
verifying that the MVT and the transaction options comply with a transaction policy, the MVT and the transaction options being verified based on the scoring;
optimizing the transaction options based on a preference; and
transmitting, via the API:
data indicating whether the transaction proposal is approved, wherein whether the transaction proposal is approved is determined based on whether the MVT complies with the transaction policy, and
at least one transaction offer based on the optimized transaction options.

14. The method of claim 13 , further comprising:
generating a second plurality of transaction options, the second plurality of transaction options being generated based on the transaction proposal and the MVT;
scoring at least one of the second plurality of transaction options, the at least one of the second plurality of transaction options being scored using the analytical model;
verifying that at least one of the second plurality of transaction options complies with the transaction policy, the at least one of the second plurality of transaction options being verified based on the scoring; and
transmitting the at least one of the second plurality of transaction options that complies with the transaction policy.

15. The method of claim 14 , wherein the at least one of the second plurality of transaction options are transmitted when the transaction proposal is rejected.

16.

Assignees

• Capital One Services Llc

Inventors

• Dhandapani Chandra
• Bajaj Raman
• Singh Gurmeet
• Karuthakantakath Ajmal
• Crable Fredrick
• Dolle Nicholas
• Repaka Vikramaditya
• Yajnik Sanjiv

Classifications

• G06Q40/03

  Credit; Loans; Processing thereof · CPC title

• G06N20/00

  Machine learning · CPC title

• G06Q10/087 Primary

  Inventory or stock management, e.g. order filling, procurement or balancing against orders · CPC title

• G06F30/20

  Design optimisation, verification or simulation (optimisation, verification or simulation of circuit designs G06F30/30) · CPC title

• G06N5/045

  Explanation of inference; Explainable artificial intelligence [XAI]; Interpretable artificial intelligence · CPC title