Fuzzy Decision Tree Based On AFS Fuzzy Logic

Learning algorithm of fuzzy decision tree bases on the instances, and focuses on deducing classification rules represented by decision tree from a group of out-of-order and out-of-rule samples. And the fuzzy decision tree has the significant theoretical meaning and practical value in the research fields of artificial intelligence, such as machine learning, data mining and intelligent control, etc.From the point of view of analyzing and comparison of the decision tree algorithms, the classical algorithms of ID3 and C4.5 are described, then the advantages and disadvantages of two algorithms are analyzed and compared. The algorithm of ID3 usually leans to select attributes which value are much more, and can't handle with continuous data, and is sensitivity to the noise. The algorithm of C4.5 is an optimized one based on the ID3 algorithm. It can process the attributes which have continuous value and empty value data. Whereas the C4.5 algorithm tends to select the attribute with minimum entropy value, not that contributes most to classification.This paper introduces a method to construct a fuzzy rule-based classifier making use of theoretical findings of the Axiomatic Fuzzy Set (AFS) theory. Compared with other fuzzy decision classifier systems, the classifier exhibits an essential advantage being of practical relevance. The reliability (relevance) of classification results is quantified by using associated confidence levels (degrees). This quantification can be applied to the data sets with mixed data type attributes. We have experimented with various data sets commonly discussed in the literature. We have also compared the obtained results with those reported for C4.5 and C-decision trees. It was found that accuracy on test data is found to be better than those produced by the other decision trees.In this paper's application, just the order relationships of the samples on the attributes are used. Experimental results demonstrate that a high accuracy can be obtained by the proposed fuzzy decision tree algorithm only according to the order relations of the attributes, instead of the numerical representations of the attributes. The proposed fuzzy decision algorithm can be applied to the data sets with various data types such as real numbers, Boolean values, partial orders, even human intuition descriptions.