US 12,591,672 B2Granted

Systems and methods for performing non-binary classification during sequence mining

AssigneeCapital One Services, LLC

InventorSamuel Sharpe

Life of this patent

Estimating the term…

Loading patent…

Abstract

Systems and methods recite a system that expands the available amount of data that may be used for training and/or detecting cyber fraud. In particular, as opposed to limiting training data (and/or successful matches) to patterns of text strings in electronic communications of user records, the systems and methods additionally generate training data (and/or successful matches) on the characteristics of the text strings themselves (e.g., the specific characters used, the length, the uniqueness of the text string, etc.). Moreover, the systems and methods may determine particular rules that an identified text string and/or characteristics thereof violates. The system may then use instances of the identified text string, characteristics thereof, and/or violated rule to supplement existing training data. Thus, the systems and methods mitigate the technical problem related to the limited training data.

Claims20

Dependent claims

A system for performing non-binary classification during sequence mining, the system comprising:

one or more processors;
and
one or more non-transitory, computer-readable mediums comprising instructions recorded thereon that when executed by the one or more processors cause operations comprising:
receiving a first user record, wherein the first user record comprises a first plurality of text strings;
parsing, using a first artificial intelligence model, the first plurality of text strings to identify a first text string in the first user record, wherein the first artificial intelligence model is trained on first training data, wherein the first training data is generated based on labeled historical text strings in a plurality of historical user records;
comparing the first text string to the first plurality of text strings to determine a first characteristic for the first text string;
parsing, using a second artificial intelligence model, the first text string to determine a second characteristic for the first text string, wherein the second artificial intelligence model is trained on second training data, wherein the second training data is generated based on labeled historical characters in a plurality of historical text strings;
comparing, using a third artificial intelligence model, the first text string to a first rule to determine a third characteristic for the first text string, wherein the third artificial intelligence model is trained on third training data, wherein the third training data is generated based on labeled rule violations in a plurality of detected cyberattacks;
normalizing outputs from the first artificial intelligence model, the second artificial intelligence model, and the third artificial intelligence model by transforming the outputs using a first classification weight based on the first characteristic, a second classification weight based on the second characteristic, and a third classification weight based on third characteristic;
determining, using a fourth artificial intelligence model, a first classification likelihood for the first text string based on the first classification weight, the second classification weight, and the third classification weight, wherein the fourth artificial intelligence model is trained on fourth training data, wherein the fourth training data is generated based on labeled indicia of detected cyberattacks;
and
generating for display, on a user interface, a recommendation based on the first classification likelihood.

9 more elements

A method for performing non-binary classification during sequence mining, the method comprising:

receiving a first user record, wherein the first user record comprises a first plurality of text strings;
parsing the first plurality of text strings to identify a first text string in the first user record;
comparing the first text string to the first plurality of text strings to determine a first characteristic for the first text string;
parsing the first text string to determine a second characteristic for the first text string;
comparing the first text string to a first rule to determine a third characteristic for the first text string;
determining a first classification weight based on the first characteristic, a second classification weight based on the second characteristic, and a third classification weight based on third characteristic;
and
determining a first classification likelihood for the first text string based on the first classification weight, the second classification weight, and the third classification weight.

5 more elements

One or more non-transitory, computer-readable mediums comprising instructions recorded thereon that when executed by one or more processors cause operations comprising:

parsing a first plurality of text strings to identify a first text string in a first user record;
comparing the first text string to the first plurality of text strings to determine a first characteristic for the first text string;
parsing the first text string to determine a second characteristic for the first text string;
comparing the first text string to a first rule to determine a third characteristic for the first text string;
determining a first classification weight based on the first characteristic, a second classification weight based on the second characteristic, and a third classification weight based on third characteristic;
and
determining a first classification likelihood for the first text string based on the first classification weight, the second classification weight, and the third classification weight.

4 more elements

Dependencies are inferred from the claim text.