Universal tree interpreter for data mining models

• US 6,941,318 B1
• Filed: 05/10/2002
• Issued: 09/06/2005
• Est. Priority Date: 05/10/2002
• Status: Active Grant

• Current Assignee: Oracle International Corporation (Oracle Corporation)
• Original Assignee: Oracle International Corporation (Oracle Corporation)
• Inventors: Tamayo, Pablo, Drescher, Gary
• Primary Examiner(s): Mizrahi, Diane D.
• Application Number:
  US10/142,169
• Time in Patent Office: 1,215 Days
• Field of Search: 707/2, 707/3, 707/6, 707/102, 707/104.1, 700/31, 345/400, 345/440, 706/45, 715/500
• US Class Current: 707/778
• CPC Class Codes: G06F 18/24323 Tree-organised classifiers
• CPC Class Codes: Y10S 707/99943 Generating database or data structure, e.g. via user interface

Abstract

A method, system, and computer program product for generating a representation of a data mining model that improves the transparency of data mining models so as to be more easily interpretable by human users. The method comprises the steps of: receiving a dataset, generating a callable version of the data mining model, and generating a tree representing decisional logic of the data mining model using the dataset.

Claims

1. A computer-implemented method of transforming a dataset to form a representation of a data mining model comprising the steps of:

receiving the dataset;

generating a callable version of the data mining model by encapsulating the data mining model in a callable wrapper that allows the data mining model to be directly called from other program code; and

generating a tree representing decisional logic of the data mining model using the dataset.

2. The computer-implemented method of claim 1, wherein the generating step comprises the steps of:

creating a split and a node using output of the callable version of the data mining model on the dataset and on created data; and

for each node, repeating the step of creating a split and a node while there is at least a minimum number of data points of the dataset in the node.

3. The computer-implemented method of claim 2, wherein the creating step comprises the steps of:

creating new data points based on a distribution of data in the dataset;

determining class membership for the created data points using the callable version of the data mining model;

determining a split location;

refining the split location; and

performing the split at the refined split location.

4. The computer-implemented method of claim 3, wherein the step of determining class membership for the created data points comprises the step of:

for each created datapoint, scoring the datapoint with the callable version of the model.

5. The computer-implemented method of claim 4, wherein the step of determining a split location comprises the step of:

looping over variables associated with both the dataset and the created data point and computing a splitting index.

6. The computer-implemented method of claim 5, wherein the splitting index is a Gini index.

7. The computer-implemented method of claim 3, wherein the representation of decisional logic of the data mining model comprises a decision tree.

8. The computer-implemented method of claim 3, wherein the decision tree comprises a classification tree.

9. A computer program product for generating a representation of a data mining model process in an electronic data processing system, comprising:

a computer readable medium;

computer program instructions, recorded on the computer readable medium, executable by a processor, for performing the steps of;

receiving a dataset;

generating a callable version of the data mining model; and

generating a tree representing decisional logic of the data mining model using the dataset.

10. The computer program product of claim 9, wherein the generating step comprises the steps of:

creating a split and a node using output of the callable version of the data mining model on the dataset and on created data; and

for each node, repeating the step of creating a split and a node while there is at least a minimum number of data points of the dataset in the node.

11. The computer program product of claim 10, wherein the creating step comprises the steps of:

creating new data points based on a distribution of data in the dataset;

determining class membership for the created data points using the callable version of the data mining model;

determining a split location;

refining the split location; and

performing the split at the refined split location.

12. The computer program product of claim 11, wherein the step of determining class membership for the created data points comprises the step of:

for each created datapoint, scoring the datapoint with the callable version of the model.

13. The computer program product of claim 12, wherein the step of determining a split location comprises the step of:

looping over variables associated with both the dataset and the created data point and computing a splitting index.

14. The computer program product of claim 13, wherein the splitting index is a Gini index.

15. The computer program product of claim 11, wherein the representation of decisional logic of the data mining model comprises a decision tree.

16. The computer program product of claim 11, wherein the decision tree comprises a classification tree.

17. A system for generating a representation of a data mining model, comprising:

a processor operable to execute computer program instructions; and

a memory operable to store computer program instructions executable by the processor, for performing the steps of;

receiving a dataset;

generating a callable version of the data mining model; and

generating a tree representing decisional logic of the data mining model using the dataset.

18. The system of claim 17, wherein the generating step comprises the steps of:

creating a split and a node using output of the callable version of the data mining model on the dataset and on created data; and

for each node, repeating the step of creating a split and a node while there is at least a minimum number of data points of the dataset in the node.

19. The system of claim 18, wherein the creating step comprises the steps of:

creating new data points based on a distribution of data in the dataset;

determining class membership for the created data points using the callable version of the data mining model;

determining a split location;

refining the split location; and

performing the split at the refined split location.

20. The system of claim 19, wherein the step of determining class membership for the created data points comprises the step of:

for each created datapoint, scoring the datapoint with the callable version of the model.

21. The system of claim 20, wherein the step of determining a split location comprises the step of:

looping over variables associated with both the dataset and the created data point and computing a splitting index.

22. The system of claim 21, wherein the splitting index is a Gini index.

23. The system of claim 19, wherein the representation of decisional logic of the data mining model comprises a decision tree.

24. The system of claim 19, wherein the decision tree comprises a classification tree.