Method, system, and computer program product for visualizing a decision-tree classifier

• US 6,278,464 B1
• Filed: 03/07/1997
• Issued: 08/21/2001
• Est. Priority Date: 03/07/1997
• Status: Expired due to Term

• Current Assignee: RPX Corporation
• Original Assignee: Silicon Graphics Incorporated (Graphics Properties Holdings, Inc.)
• Inventors: Tesler, Joel D., Kohavi, Ron
• Primary Examiner(s): Nguyen, Phu K.
• Application Number:
  US08/813,336
• Time in Patent Office: 1,628 Days
• Field of Search: 345/440, 345/441, 345/442, 345/357
• US Class Current: 345/440
• CPC Class Codes: G06F 18/40 Software arrangements specially adapted for pattern recognition, e.g. user interfaces or toolboxes therefor
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06Q 10/10 Office automation; Time management

Abstract

A method, system and a computer program product for visualizing a decision-tree classifier are provided. The structure of a decision-tree classifier is mapped into a three-dimensional decision-tree visualization. The three-dimensional decision-tree visualization is displayed, representing the structure of the decision-tree classifier. Nodes in the three-dimensional decision-tree visualization include graphical objects representing nodes arranged in a hierarchy. The nodes in the three-dimensional decision-tree visualization correspond to nodes in the decision-tree classifier. Graphical attributes representative of information at corresponding nodes of the decision-tree classifier are provided at each node in the three-dimensional decision-tree visualization.

Claims

1. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents the number of records in a class at each node, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of the number of records in a class at the corresponding node of the decision-tree classifier.

2. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents a distribution of records at each node, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of the distribution of records at the corresponding node of the decision-tree classifier.

3. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents purity of classification at each node, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of the purity of classification at the corresponding node of the decision-tree classifier.

4. The method of claim 3, wherein said purity of classification indicates how easy it is to make a prediction as to the class of a record, wherein said step (a) comprises the step of:

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute indicating how easy it is to make a prediction as to the class of a record at the corresponding node of the decision-tree classifier.

7. The method of claim 3, wherein said providing step comprises the step of:

computing purity by an entropic formula.

5. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents estimated accuracy of classification at each node, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of the estimated accuracy of classification at the corresponding node of the decision-tree classifier.

6. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents estimated reliability of classification at each node, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of the estimated reliability of classification at the corresponding node of the decision-tree classifier.

8. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier, wherein a first graphical attribute comprises a bar, each said node having at least one bar, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one bar representative of information at the corresponding node of the decision-tree classifier.

9. The method of claim 8, wherein each bar represents a class, wherein step (a) comprises the step of:

at each node in the three-dimensional decision-tree visualization, providing at least one bar representative of a class at the corresponding node of the decision-tree classifier.

10. The method of claim 8, wherein the height of each bar represents the number of records in each class, wherein step (a) comprises the step of:

at each node in the three-dimensional decision-tree visualization, providing at least one bar with a height representative of the number of records in a class at the corresponding node of the decision-tree classifier.

11. The decision-tree visualizer according to claim 8, wherein the number of records in each bar indicate the estimated reliability of the classification at each node, wherein step (a) comprises the step of:

at each node in the three-dimensional decision-tree visualization, providing at least one bar with a number of records indicating the estimated reliability of the classification of the corresponding node of the decision-tree classifier.

12. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier wherein a second graphical attribute is a base, each said node has a representative base, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one base representative of information at the corresponding node of the decision-tree classifier, wherein each base has a color that indicates the purity of the classification, wherein step (a) comprises the step of;

(i) at each node in the three-dimensional decision-tree visualization, providing at least one base indicating the purity of the classification at the corresponding node of the decision-tree classifier.

14. The method of claim 12, wherein the color red indicates low purity, yellow indicates fair purity and green indicates high purity, wherein step (i) comprises the step of:

at each node in the three-dimensional decision-tree visualization, providing at least one base with a color indicating the level of purity at the corresponding node of the decision-tree classifier.

15. The method of claim 12, wherein said providing step comprises the step of:

computing purity by an entropic formula.

13. A computer-implemented method for visualizing a decision-tree classifier, comprising the steps of:

mapping a structure of the decision-tree classifier into a three-dimensional decision-tree visualization, wherein the decision-tree classifier is induced from a training set, and wherein the three-dimensional decision-tree visualization represents the structure of the decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier;

displaying the three-dimensional decision-tree visualization; and

at each node in the three-dimensional decision-tree visualization, providing at least one graphical attribute representative of information at a corresponding node of the decision-tree classifier wherein a second graphical attribute is a base, each said node has a representative base, wherein said providing step comprises the step of;

(a) at each node in the three-dimensional decision-tree visualization, providing at least one base representative of information at the corresponding node of the decision-tree classifier, wherein each base has a height that represents the total number of records at each node, wherein step (a) comprises the step of;

at each node in the three-dimensional decision-tree visualization, providing at least one base with a height representative of the total number of records at the corresponding node of the decision-tree classifier.

16. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute represents the number of record in a class at each node.

17. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute represents a distribution of records at each node.

18. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute represents purity of classification at each node.

19. The system of claim 18, wherein said purity of classification indicates how easy it is to make a prediction as to the class of a record.

21. The system of claim 18, wherein said purity is computed by an entropic formula.

20. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute represents estimated reliability of classification at each node.

22. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute is a bar, each said node having at least one of said bar.

23. The system of claim 22, wherein each bar represents a class.

24. The system of claim 22, wherein the height of each bar represents the number of records in each class.

25. The system of claim 24, wherein the number of records in each bar indicate the estimated reliability of the classification at each node.

26. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at lease one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute is a base, each said node having at least one said base, wherein each base has a color that indicates the estimated accuracy of the classification at each node.

27. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at lease one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute is a base, each said node having at least one said base, wherein each base has a color that indicates the purity of classification at each node.

29. The system of claim 27, wherein the color red indicates low purity, yellow indicates fair purity and green indicates high purity.

28. A computer-implemented system for visualizing a decision-tree classifier, comprising:

means for mapping a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

means for displaying said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

at each node in said three-dimensional decision-tree visualization, means for providing at least one graphical object, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute is a base, each said node having at least one said base, wherein each base has a height that represents the total number of records at each node.

30. A computer program product comprising a computer useable medium having computer program logic recorded thereon for enabling a processor in a computer system to visualize a decision-tree classifier, comprising:

(a) means for enabling said processor to map a structure of said decision-tree classifier into a three-dimensional decision-tree visualization having nodes and lines, wherein said decision-tree classifier is induced from a training set, and wherein said three-dimensional decision-tree visualization represents said structure of said decision-tree classifier;

(b) means for enabling said processor to cause the displaying of said three-dimensional decision-tree visualization, wherein said nodes in said three-dimensional decision-tree visualization are arranged in a hierarchy and connected by said lines, said nodes in said three-dimensional decision-tree visualization correspond to nodes in said decision-tree classifier; and

(c) means for enabling a processor to provide at least one graphical object at each node in said three-dimensional decision-tree visualization, said graphical object having at least one graphical attribute representative of information at a corresponding node of said decision-tree classifier, wherein a graphical attribute represents the number of record in a class at each node.

31. A decision-tree classifier visualization system, comprising:

an inducer that generates at least one visualization file representative of a three-dimensional visualization of a structure of a decision-tree classifier, wherein said inducer generates said at least one visualization file based on a training set;

wherein said at least one visualization file can be transformed into a three-dimensional visualization of a decision-tree classifier, and wherein said three-dimensional decision-tree classifier visualization represents said structure of said decision-tree classifier, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier, wherein at each node in the three-dimensional decision-tree visualization at least one graphical attribute is representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents the number of records in a class at each node.

32. The system of claim 31, wherein said inducer includes a mapping module that maps a training set used by the inducer to generate the decision-tree classifier to said at least one visualization file.

33. A network system, comprising:

a tool manager that generates a configuration file;

a data mover that extracts data from a data source corresponding to said configuration file;

a data miner that receives extracted data from said data mover and generates a data file, wherein said data file is representative of a structure of a decision-tree classifier induced from a training set;

a data visualization tool that receives said data file from said data miner and generates a three-dimensional decision-tree classifier visualization, wherein said three-dimensional decision-tree classifier visualization represents said structure of said decision-tree classifier; and

a display that displays said three-dimensional decision-tree classifier visualization, wherein the three-dimensional decision-tree visualization includes nodes, the nodes in the three-dimensional decision-tree visualization including graphical objects representing nodes arranged in a hierarchy, the nodes in the three-dimensional decision-tree visualization corresponding to nodes in the decision-tree classifier, wherein at each node in the three-dimensional decision-tree visualization at least one graphical attribute is provided that is representative of information at a corresponding node of the decision-tree classifier, wherein a graphical attribute represents the number of records in a class at each node.

34. The system of claim 33, wherein said configuration file identifies the content to be displayed in a data visualization and scope of external dimension space.