Method, system and program product for developing a data model in a data mining system

• US 7,367,011 B2
• Filed: 04/13/2004
• Issued: 04/29/2008
• Est. Priority Date: 04/13/2004
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: International Business Machines Corporation
• Inventors: Selby, David A., Ramsey, Mark S.
• Primary Examiner(s): Zhen; Wei
• Assistant Examiner(s): RAMPURIA, SATISH
• Application Number:
  US10/825,404
• Publication Number:
  US 20050228767A1
• Time in Patent Office: 1,477 Days
• Field of Search: 717104-119, 707100-200
• US Class Current: 717/104
• CPC Class Codes: G06F 16/283 Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A database table of predefined data transformations is provided. Each predefined data transformation is associated in the table with a unique identifier, a corresponding description and a validity period. When a data modeler wishes to develop a data model for a desired prediction, he/she will first determine a set of variables that will be used therefor. The set of variables can include any of the predefined data transformations from the database table. The data model will then be developed by applying raw data to the set of variables and determining a mathematical relationship there between. Once the data model has been developed, the data modeler will write a reusable specification for applying the data model operationally. Thereafter, IT personnel or the like can code and deploy the data model using the specification.

Claims

1. A method for developing a data model in a data mining system, comprising:

providing a database table of user predefined data transformations, the data

transformations in the database table being reusable by a plurality of data models;

providing raw data from a data mining operation;

developing a data model of variables using at least one data transformation selected from the database table and the raw data; and

writing a specification for applying the data model operationally, wherein the step of writing a specification comprises writing a reusable set of instructions for applying the data model operationally,

wherein the database table of predefined data transformations associates each of the predefined data transformations with a unique identifier, a description and a validity period and wherein the developing step comprises retrieving the at least one predefined data transformation from the database table according to its unique identifier;

wherein the developing step includes determining a set of variables for a desired prediction, wherein the set of variables include at least one predefined data transformation selected from the database table; and

deriving a mathematical relationship between the set of variables.

2. The method of claim 1, farther comprising coding and deploying the data model using the specification.

3. The method of claim 1, wherein the developing step further comprises applying the raw data to the set of variables.

4. The method of claim 1, wherein the step of providing raw data comprises extracting raw data from a data warehouse.

5. The system of claim 1, further comprising a modification policy that governs modification of the predefined data transformations in the database table.