Behavioral Misalignment Detection Within Entity Hard Segmentation Utilizing Archetype-Clustering

Patent metadata

Field	Value
Publication number	US-2017270428-A1
Application number	US-201615074856-A
Country	US
Kind code	A1
Filing date	Mar 18, 2016
Priority date	Mar 18, 2016
Publication date	Sep 21, 2017
Grant date	—

Abstract

An automated way of learning archetypes which capture many aspects of entity behavior, and assigning entities to a mixture of archetypes, such that each entity is represented as a distribution across multiple archetypes. Given those representations in archetypes, anomalous behavior can be detected by finding misalignment with a plurality of entities archetype clustering within a hard segmentation. Extensions to sequence modeling are also discussed. Applications of this method include anti-money laundering (where the entities can be customers and accounts, as described extensively below), retail banking fraud detection, network security, and general anomaly detection.

First claim

What is claimed is:

1 . A method to be performed by a computer processor, the computer processor forming at least part of a computer system, the method comprising:
maintaining one or more profiles in a data store for a plurality of entities of interest, each of the one or more profiles being formed as a data structure that captures statistics of one or more behaviors of an entity associated with the profile, the data structure further including demographic information associated with the entity; generating one or more models based on the captured statistics of one or more behaviors of the plurality of entities, the one or more models being used by the computer processor for predicting a behavior of a new entity of interest; assigning the plurality of entities of interest to hard segments of a segmentation scheme; and, assigning individual ones of the plurality of entities of interest to a set of archetypes, the set of archetypes being an archetype distribution and generated based on the generated one or more models, each archetype of the set of archetypes indicating at least one behavior characteristic that the entities assigned to that archetype have in common.

2 . The method of claim 1 , further comprising:
identifying a transaction performed by an entity, of the plurality of entities of interest, wherein, the at least one behavior characteristic of the entity produces an archetype distribution to which the entity is assigned; determining a variation over time between the archetype distribution to which the entity, of the plurality of entities of interest, is assigned and a distance from the set of archetypes soft clusters associated with other entities in the hard segmentation to which the entity is assigned; generating a soft clustering misalignment score based on the determined variation, the soft clustering misalignment score indicating the degree of variation between the archetype distribution to which the entity is assigned and the set of archetypes soft clusters associated with the other entities in the hard segmentation to which the entity is assigned; and, generating an alert in response to identifying that the determined soft clustering misalignment score for the entity exceeds a threshold.

3 . The method of claim 2 , further compromising:
generating a report indicating entities having a soft-clustering misalignment score that is indicative of a need for further anti-money laundering investigation.

4 . The method of claim 2 , further compromising:
generating a report indicating entities having a soft cluster misalignment score that exceeds a threshold indicative that the entity needs to be reassigned to a different hard segment.

5 . The method of claim 2 , further compromising:
identifying an entity that has a soft cluster misalignment score indicative that the entity has migrated over time from a first soft cluster of archetypes to a second soft cluster of archetypes; and, generating a report indicating that the identified entity exhibits sleeper behaviors or radicalization.

6 . The method of claim 1 , further comprising:
assigning the plurality of entities to a set of archetypes, the set of archetypes being a archetype distribution, based on entity transaction behavior information and entity demographic information.

7 . The method of claim 1 , further comprising:
receiving one or more streams of data associated with the plurality of entities and wherein the models are generated based on the received streams of data.

8 . The method of claim 7 , wherein the one or more streams of data include transaction data associated with the plurality of entities of interest.

9 . The method of claim 7 , wherein the one or more streams of data include demographic markers associated with the plurality of entities of interest.

10 . The method of claim 7 , further comprising:
updating the at least one behavior characteristic of at least one archetype of the set of archetypes, based on an update to the models caused by the received one or more streams of data associated with the plurality of entities.

11 . The method of claim 1 , wherein the one or more models are configured to translate the captured recursive variables and statistical behavior to the set of archetypes.

12 . The method of claim 1 , wherein each archetype of the set of archetypes, to which an entity is assigned, indicates a behavior of the entity.

13 . The method of claim 1 , wherein each entity of the plurality of entities of interest are assigned to multiple archetypes of the set of archetypes.

14 . The method of claim 1 , wherein assigning individual entities of the plurality of entities of interest to a set of archetypes comprises:
utilizing, collaborative profiling of behavior events associated with the plurality of entities of interest, the behavior events associated with financial transactions performed by the plurality of entities; and, utilizing, collaborative profiling of behavior events associated with the plurality of entities of interest, the behavior events associated with demographic data associated with the plurality of entities.

15 . The method of claim 14 , further comprising:
capturing behavior events by using recurrent networks technology to capture sequences of behavior events.

16 . A system comprising:
at least one programmable processor; and a machine-readable medium storing instructions that, when executed by the at least one processor, cause the at least one programmable processor to perform operations comprising:
maintaining one or more profiles in a data store for a plurality of entities of interest, each of the one or more profiles being formed as a data structure that captures statistics of one or more behaviors of an entity associated with the profile, the data structure further including demographic information associated with the entity;
generating one or more models based on the captured statistics of one or more behaviors of the plurality of entities, the one or more models being used by the computer processor for predicting a behavior of a new entity of interest;
assigning the plurality of entities of interest to hard segments of a segmentation scheme; and,
assigning individual ones of the plurality of entities of interest to a set of archetypes, the set of archetypes being an archetype distribution and generated based on the generated one or more models, each archetype of the set of archetypes indicating at least one behavior characteristic that the entities assigned to that archetype have in common.

17 . The system of claim 16 , wherein the operations further comprise:
identifying a transaction performed by an entity, of the plurality of entities of interest, wherein, the at least one behavior characteristic of the entity produces an archetype distribution to which the entity is assigned; determining a variation over time between the archetype distribution to which the entity, of the plurality of entities of interest, is assigned and a distance from the set of archetypes soft clusters associated with other entities in the hard segmentation to which the entity is assigned; generating a soft clustering misalignment score based on the determined variation, the soft clustering misalignment score indicating the degree of variation between the archetype distribution to which the entity is assigned and the set of archetypes soft clusters associated with the other entities in the hard segmentation to which the entity is assigned; and, generating an alert in response to identifying that the determined soft clustering mi

Assignees

• Fair Isaac Corp

Inventors

• Zoldi Scott Michael
• Murray Joseph F

Classifications

• G06Q10/10

  Office automation; Time management · CPC title

• G06Q30/06

  Buying, selling or leasing transactions · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N7/005

  Physics · mapped topic

• G06N99/005 Primary

  Physics · mapped topic