System and method for supporting the utilization of machine language

• US 7,809,663 B1
• Filed: 05/22/2007
• Issued: 10/05/2010
• Est. Priority Date: 05/22/2006
• Status: Active Grant

• Current Assignee: Concentrix CVG Customer Management Delaware LLC (Synnex Corporation)
• Original Assignee: Convergys Corporation
• Inventors: Irwin, Scott A., Craig, Brian D., Weagraff, Stephen D., Birch, Robert D.
• Primary Examiner(s): Sparks; Donald
• Assistant Examiner(s): Fernandez Rivas; Omar F
• Application Number:
  US11/751,680
• Time in Patent Office: 1,232 Days
• Field of Search: 706/12, 706 15- 21, 706 45- 47, 706/59, 706/62, 707 1- 3, 707/6, 707/100, 715/700, 715/701, 715/707, 715/738, 715/744
• US Class Current: 706/45
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A system and method is disclosed which integrates a machine learning solution into a large scale, distributed transaction processing system using a supporting architecture comprising a combination of computer hardware and software. Methods of using a system comprising such supporting architecture provide application designers access to the functionality included in a machine learning solution, but might also provide additional functionality not supported by the machine learning solution itself.

Claims

1. A computerized method of incorporating a machine learning solution, comprising a machine learning model, into a transaction processing system, the method comprising:

a) configuring an application interface component to access a set of functionality associated with the machine learning solution;

b) configuring a model management component to access at least one instance of said machine learning model according to a request received from the application interface component, wherein configuring said model management component comprises the steps of;

configuring a synchronization policy associated with said model management component;

configuring a persistence policy associated with said model management component; and

configuring a versioning policy associated with said model management component;

wherein said versioning policy is implemented according to a versioning policy interface which encodes how a machine learning model version in memory and a machine learning model version in persistent storage should be synchronized.

2. The method of incorporating said machine learning solution, as claimed in claim 1, wherein said persistence policy comprises a set of computer-executable instructions encoding how, when, and where said machine learning model should be persisted.

3. The method of incorporating said machine learning solution, as claimed in claim 1, further comprising accessing said synchronization policy through a synchronization policy interface.

4. The method of incorporating said machine learning solution, as claimed in claim 3, wherein said synchronization policy interface comprises computer-executable instructions for invoking said synchronization policy for said machine learning model.

5. A machine learning system comprising:

a computer comprising;

a) an application interface component further comprising;

a machine learning engine;

a machine learning context; and

a machine learning controller;

b) a model management component further comprising;

a model pool;

a model pool manager; and

a synchronization manager;

c) an algorithm management component further comprising;

an algorithm manager;

an algorithm instance map;

an algorithm instance map;

an algorithm factory;

an algorithm implementation; and

an algorithm interface;

wherein the machine learning engine, in combination with the machine learning context, provides said application interface component, accessible by an application program, and wherein said machine learning engine communicates a request to said algorithm management component and/or said model management component;

wherein the algorithm interface defines how an algorithm instance of an algorithm implementation is accessed;

wherein the algorithm manager is configured to create, retrieve, update and/or delete an algorithm instance, through said algorithm factory, and enter said algorithm instance into said algorithm instance map based on said request received through said application interface component;

wherein a model is stored within said model pool and wherein a synchronization policy is associated with said model;

wherein said model pool manager is configured to create, retrieve, update and/or delete said model within the model pool based on said request received through said application interface component;

wherein said synchronization manager executes said synchronization policy associated with said model;

wherein said machine learning controller binds an algorithm instance with said model;

wherein said synchronization manager is configured to update a prototypical model only if said request contains an appropriate learning event; and

wherein said appropriate learning event comprises that said model, associated with said request, and said prototypical model comprise an identical vendor, an identical algorithm structure, and a set of identical model parameter attribute types.

6. The machine learning system, as claimed in claim 5, wherein said application interface component, provided by the machine learning context, is configured to expose a particular machine learning algorithm.

7. The machine learning system, as claimed in claim 5, wherein said application interface component, provided by the machine learning context, is configured to expose a general purpose interface.

8. The machine learning system, as claimed in claim 5, wherein said application interface component, provided by the machine learning context, is configured to expose a focused interface to be applied to a specific task.

9. The machine learning system, as claimed in claim 5, wherein said synchronization manager is configured to propagate an updated prototypical model to a plurality of models comprising said model and at least one model hosted remotely from said updated prototypical model.

10. The machine learning system, as claimed in claim 9, wherein said propagation of said updated prototypical model to said plurality of models is executed in accordance with a synchronization policy associated with each of said plurality of models.

11. The machine learning system, as claimed in claim 5, wherein said model pool manager is configured to retrieve a plurality of models within the model pool based on said request received through said application interface component.

12. A method of utilizing a computerized machine learning solution, comprising a machine learning model, in a transaction processing system, the method comprising:

a) using a machine learning engine of an application interface component to connect said machine learning solution to an on-line retailer'"'"'s website;

b) coding a classification method, in said transaction processing system, to be invoked when a customer visits the website;

c) coding said classification method to send a message to a machine learning controller;

d) requesting a model manager and an algorithm manager from said machine learning controller for an instance of an associated algorithm and an instance of an associated model;

e) determining a recommendation from a set of available products based on processing said instance of said associated algorithm, said instance of said associated model and an output from said classification method;

f) coding a learn method to be called when a purchase takes place;

g) passing a purchase message, regarding said purchase, to said machine learning controller as a learning event;

h) passing said learning event to an update method exposed by a synchronization policy interface of said instance of said associated model via said model manager;

i) sending said learning event to a prototypical model;

j) creating an updated version of said prototypical model; and

k) propagating said updated prototypical model to a plurality of distributed servers according to a propagation method exposed by said synchronization policy interface.

13. The method of utilizing said machine learning solution, as claimed in claim 12, wherein said propagation method is a master-slave synchronization policy.