| US 7,533,095 B2 | ||
| Data mining within a message handling system | ||
| Mark S. Ramsey, Colleyville, Tex. (US); and David A. Selby, Nr Fareham (United Kingdom) | ||
| Assigned to International Business Machines Corporation, Armonk, N.Y. (US) | ||
| Filed on Apr. 19, 2005, as Appl. No. 11/109,446. | ||
| Prior Publication US 2006/0248033 A1, Nov. 02, 2006 | ||
| Int. Cl. G06F 7/00 (2006.01) | ||
| U.S. Cl. 707—6 [707/102] | 11 Claims |
| 6. A data mining method adapted to be performed in a computer system having a message handling system therein, said message
handling system comprising a message broker and a plurality of queues that include computer-readable memory, said method comprising:
receiving information in at least one queue of the plurality of queues, said information being received from at least one
publisher;
executing a data mining algorithm by operating upon a data mining model that depends on model parameters and on data values
for independent variables, said information comprising input data that includes the data values and further includes the model
parameters, model content, or combinations thereof, said model content being the data mining model or an identifier thereof
or a pointer thereto, said executing comprising utilizing the information to generate at least one score;
executing a decision algorithm to apply at least one rule to the at least one score to generate at least one decision;
publishing at least one result in a result queue of the plurality of queues, said at least one result being selected from
the group consisting of the at least one score, the at least one decision, and combinations thereof, said result queue being
subscribed to by at least one subscriber; and
transmitting the at least one result from the result queue to the at least one subscriber,
wherein said receiving, said executing the data mining algorithm, said executing the decision algorithm, said publishing,
and said transmitting are performed by the message broker,
wherein the data mining model is a linear or nonlinear regression model such that the model parameters comprise weights respectively
associated with the independent variables,
wherein a positive integer N is a total number of said independent variables denoted as X1, X2, . . . , XN;
wherein said weights consist of N weights denoted as W1, W2, . . . , WN;
wherein the data mining model is said nonlinear regression model comprising a function F(X1, X2, . . . , XN) expressed in terms of a nonlinear regression equation having a form of:
F(X1, X2, . . . , XN)=W1f1(X1)+W2f2(X2)+ . . . +Wnfn(XN);
wherein f1(X1), f2(X2), . . . fN(XN) are nonlinear functions of X1, X2, . . . XN, respectively.
|