Method for generating predictive models for a business problem via supervised learning

1. A method for solving a business problem, comprising:

pooling, in a data storage unit configured as a relational memory, transaction data received from a plurality of subscribers over a network, the transaction data including samples of fraudulent and non-fraudulent transactions, the samples of fraudulent transactions include a variety of fraud patterns from across different industries;

inputting the pooled transaction data into a processor configured to execute a data mining algorithm, when the data mining algorithm is executed the processor outputting a predictive model that detects a fraudulent transaction; and

providing a copy of the predictive model to the subscribers in exchange for a fee.

2. The method of claim 1 further comprising:

enrolling the subscribers.

3. The method of claim 1 wherein the transaction data complies with a normative data format.

4. A method for solving a business problem, comprising:

offering subscriptions to business entities, with a first type of subscription being offered to a first subscriber at a first subscription fee in exchange for contribution of a first quantity of transaction data from the first subscriber, and a second type of subscription being offered to a second subscriber at a second subscription fee in exchange for contribution of a second quantity of transaction data from the second subscriber, wherein the first subscription fee is higher than the second subscription fee and the first quantity of transaction data is greater than the second quantity of transaction data, the transaction data including samples of fraudulent and non-fraudulent transactions, the samples of fraudulent transactions include a variety of fraud patterns from across different industries;

receiving, at a computer, the transaction data via a network portal, the transaction data being pooled in a data storage unit configured as a relational memory, the data storage unit being associated with the computer;

inputting the pooled transaction data into the computer, the computer being configured to execute a data mining algorithm, when the computer executes the data mining algorithm the computer outputting a predictive model useful in detecting a fraudulent transaction;

providing a copy of the predictive model to the subscribers;

periodically updating the predictive model by re-executing the data mining algorithm on the computer using new pooled transaction data, the new pooled transaction data including additional transaction data received from the subscribers that includes new samples of fraudulent and non-fraudulent transactions; and

providing a copy of the updated predictive model to the subscribers.

5. The method of claim 4 wherein the transaction data complies with a normative data format.

6. A computer-readable memory encoded with a computer program for a business service provider, when executed, the computer program operable to:

register as subscribers, business entities from across different industries;

populate a data storage unit configured as a relational memory with transaction data contributed by the subscribers, the transaction data including samples of fraudulent and non-fraudulent transactions;

inputting the transaction data into a computer configured to execute a data mining algorithm, when the data mining algorithm is executed, the computer generating a predictive model that detects one or more hidden patterns in the fraudulent transactions; and

provide a copy of the predictive model to the subscribers.

7. The computer-readable memory of claim 6 further comprising:

preparing the transaction data in accordance with a normative data format.

8. The computer-readable memory of claim 6 wherein the predictive model if provided to the subscribers for a fee.

9. The computer-readable memory of claim 6 further comprising:

restricting access to the predictive model by a subscriber based on a number of fraudulent transactions contributed by the subscriber.