Neural network based decision processor and method

• US 6,782,375 B2
• Filed: 01/16/2001
• Issued: 08/24/2004
• Est. Priority Date: 01/16/2001
• Status: Active Grant

• Current Assignee: JP Morgan Chase Bank, N.A. (JP Morgan Chase & Co)
• Original Assignee: Providian Bancorp Services
• Inventors: Osman, Anas, Usman, Azhar M., Abdel-Moneim, Eiman, Mosier, Michael T.
• Primary Examiner(s): STARKS, WILBERT L
• Application Number:
  US09/761,328
• Publication Number:
  US 20020138455A1
• Time in Patent Office: 1,316 Days
• Field of Search: 706/16, 706/46, 706/52, 706/15, 706/26
• US Class Current: 706/46
• CPC Class Codes: G06N 3/02 Neural networks

Abstract

A computer network-based customer acquisition server and method of selecting preferred products includes using a neural network-based decision engine that automatically generate queries and select preferred products as a function of responses to the queries.

Claims

1. A method of selecting a preferred product from a list that includes a plurality of products, the method comprising:

(a) presenting a sequence of queries to a user, the user responding to each query in sequence with a corresponding response;

(b) in response to a user response to a query in the sequence, using a neural network to delete at least one product from the list to provide an updated list;

(c) in the event that the user does not select a preferred product from the updated list, continuing with the presentation of the sequence of queries to the user; and

(d) repeating steps (b) and (c) until the user selects a preferred product.

2. A method of selecting a preferred from a list that includes a plurality of products, the method comprising:

(a) providing a first query to a user;

(b) in response to a user response to the first query, using a neural network to delete at least one product from the list to provide an updated list;

(c) in the event that the user does not select a preferred product from the updated list, presenting a new query to the user and, in response to a user response to the new query, deleting at least one additional product from the updated list to provide a new updated list; and

(d) repeating item (c) until the user selects a preferred product.

3. The method of claim 2, and wherein the products are selected from a group consisting of banking services, mortgage services, brokerage services, credit card services, insurance services, telecommunications services, and combinations thereof.

4. The method of claim 2, and wherein the products comprise services.

5. The method of claim 2, and wherein the queries are presented to the user via a network based browser.

6. A customer acquisition server that dynamically generates an updated list of products from an original list of a plurality of products, comprising:

a neural network-based decision engine that iteratively responds to user responses to a sequence of queries provided to a user by, upon receipt of a user response to a query, deleting at least one product from a current list of products to provide the updated list of products until the user selects a preferred product from the list of products.

7. The method of claim 4, and wherein the services include stock brokerage services.

8. The method of claim 4, and wherein the services include banking services.

9. The method of claim 4, and wherein the services include insurance services.

10. The method of claim 5, and wherein the network is an internet network.

11. The method of claim 5, and wherein the network is an intranet network.

12. A customer acquisition server, comprising:

a neural network based decision engine that automatically generates a sequence of a plurality of queries and selects a preferred product as a function of responses to the queries,

wherein the decision engine generates subsequent queries as a function of a response to a previous query,

and wherein the decision engine selects the preferred product as a function of a number of response that is less than or equal to the plurality of queries.

13. The server of claim 12, and further comprising:

a database operatively connected to the decision engine, wherein the database stores the queries and query correlations.

14. The server of claim 12, and wherein the decision engine calculates a confidence level after receiving a response and wherein the decision engine selects a preferred product as a function of the responses received before the confidence level exceeds a confidence threshold.

15. The server of claim 14, and wherein the decision engine determines a confidence level after receiving each response.

16. The server of claim 12, and wherein the queries include an investor type query that classifies a user by the user'"'"'s investing preferences.

17. The server of claim 16, and wherein the investing preferences is selected from a group consisting of life goal planner, serious investor, hyperactive investor, one-stop shopper, short-term trader, long-term investor, futures trader, options trader, and combinations thereof.

18. The server of claim 12, and wherein the queries include a banking type query that classifies a user by the user'"'"'s banking preferences.

19. The server of claim 12, and wherein the queries include an insurance type query that classifies a user by the user'"'"'s insurance preferences.

20. A customer acquisition server, comprising:

a neural network based decision engine that automatically generates queries and selects a preferred product as a function of responses to the queries,

wherein the decision engine generates first and second queries, where the second query is a function of a response to the first query,

and wherein the decision engine comprises a plurality of self organizing maps comprising a plurality of memory locations for storing subwords, wherein the memory locations comprises a corresponding link weight,

and wherein the decision engine compares a response to a subset of the subwords.

21. The server of claim 20, and further comprising:

a training module for training the decision engine.

22. The server of claim 21, and wherein the training module allows iterative training of the decision engine.

23. The server of claim 20, and wherein the decision engine generates the queries using stochastic modulation.

24. The server of claim 20, and further comprising:

a retraining module that adjusts links between memory locations as a function of a response.

25. The server of claim 24, and wherein the response being responsive to a query relating to the accuracy of the neural network in selecting the preferred product.

26. The server of claim 14, and further comprising:

a retraining module that adjusts a correlation as a function of a response.

27. The server of claim 26, and wherein the retraining module adjusts a correlation as a function of an activity with the preferred product.

28. The server of claim 26, and wherein the retraining module adjusts a correlation as a function of vendor input.

29. The server of claim 28, and wherein the vendor is a vendor of stock brokerage services.

30. The server of claim 28, and wherein the vendor is a vendor of banking services.

31. The server of claim 28, and wherein the vendor is a vendor of insurance services.

32. A method of automatically selecting a preferred product from a list of products, the method comprising:

generating a plurality of queries, and

automatically selecting a preferred product from the list of products using a neural network decision engine to analyze responses to the queries,

wherein the decision engine automatically generates first and second queries, where the second query is a function of a response to the first query,

and wherein the decision engine selects the preferred product is a function of fewer responses than the plurality of queries.

33. The method of claim 32, and further comprising:

presenting the queries to a user via a network based browser.

34. The method of claim 33, and wherein the network is an internet network.

35. The method of claim 33, and wherein the network is an intranet network.

36. The method of claim 32, and wherein the step of generating queries comprises generating queries using stochastic modulation.

37. The method of claim 32, and wherein the step of generating queries comprises generating a retraining query and retraining the neural network as a function of a response to the retraining query.

38. The method of claim 32, and wherein the step of generating queries comprises generating a query with a set of response options, where the response options are a function of trend data.

39. The method of claim 38, and wherein the trend data includes stock-related trend data.

40. The method of claim 38, and wherein the trend data includes banking-related trend data.

41. The method of claim 38, and wherein the trend data includes insurance-related trend data.

42. The method of claim 38, and wherein the trend data comprises market trend data.

43. The method of claim 38, and wherein the trend data comprises user trend data.

44. The method of claim 32, and wherein the step of selecting a preferred product comprises selecting a preferred product as a function of a user profile that includes information submitted during a current user session.

45. The method of claim 44, and wherein the user profile includes information relating to responses from a user'"'"'s prior selection process.

46. The method of claim 44, and wherein the user profile comprises information relating to responses from a user'"'"'s prior activities with a vendor.

47. The method of claim 44, and further comprising:

generating a confidence level as a function of the responses, wherein the generating queries includes generating queries until the confidence level exceeds a confidence threshold.

48. The method of claim 32, and further comprising:

retraining the neural network.

49. The method of claim 48, and wherein the retraining the neural network comprises retraining the neural network as a function of the responses.

50. The method of claim 48, and wherein the retraining the neural network comprises retraining the neural network as a function of response trends.

51. The method of claim 48, and wherein the retraining the neural network comprises retraining the neural network as a function of market trends.