Fuzzy logic extension of nonparametric approach to feature selection and binary decision tree design

The binary decision tree for pattern recognition systems based on the selection of a set of features is both capable of high discrimination and is economical. A further extension is proposed which will incorporate fuzzy logic decisions at each node as part of the feature selection process. Each node will return a possibilistic response: “possibly” or “possibly not” as the decision regarding a set p.f. patterns. The advantage to fuzzy decision nodes is that the input to the decision tree may be linguistic in nature and the tree could return a `crisp´ or defuzzified output, which will be a useful result