Re: DM: Fuzzy Trees

[Mark Last]

Hi Alberto,

On fuzzy decision trees, please see the following paper:

Yuan and Shaw, Induction of Fuzzy Decision Trees, Fuzzy Sets and 
Systems 69 (1995) 125-139

In the above paper, the rule accuracy is measured by "fuzzy 
subsethood" and each predicted class is assigned a classification
truth level.

Mark


At 02:04 PM 10/5/99 -0300, alberto wrote:
>I am working with a Fuzzy Binary (Low-High) Tree model for
>classification and DataMining, and I have some doubts when I
>try to apply the "crisp" criteria of Accuracy and Coverage. Does 
>anyone
>Know a 'Fuzzy' criteria to evaluate the Fuzzy rules
>(i ex.:  If X1 is Low and X2 is Low and X3 then Class= * ;  Accuracy
>(?)  / Coverage (?))
>
>***** FUZZY Binary (Low-High) TREE ******
>
>                  x1-[L-H]
>                        /     \
>         x2- [L-H]  x4-[L-H]
>               /      \         /      \
>   x3-[L-H]  [L-H]   *      [L-H]
>          /   \     /   \               /    \
>        *    *  *    *             *    *
>
>And I also have problems to determinate the class at each final node,
>because of the Fuzzy partitioning. In fact each patter is
>not only present in one final node, it is present in several nodes 
>with
>a different 'degree' (alpha). Then how to say if that rule-i
>classifies class 1, 2 or...n??
>
>Any help or suggestions will be welcome.
>Thank You very much
>
>Alberto Iriarte
>PUC-Rio University
>Rio de Janeiro
>
>
---------------------------------------------
Mark Last
Visiting Assistant Professor
Computer Science and Engineering
University of South Florida
4202 E. Fowler Ave., ENB 118
Tampa, FL 33620, USA
Tel: 813/974-4763
Fax: 813/974-5456
mlast@csee.usf.edu
WWW: http://www.csee.usf.edu/~mlast/