Classification based on stable fuzzy decision tree method

Abstract

Decision tree induction is one of useful approaches for extracting classification knowledge from set instances. Considerable part of these instances obtains from formal analysis and modeling of human activities, which has fuzzy nature. It is often the case that real-world tasks can be handled easily by humans, they are often too difficult to be handled by machines. Fuzzy logic allows us to describe this problem. Fuzzy decision tree is a very popular method for fuzzy classification. We introduced term of cumulative information estimations based on Theory of Information approach. We used these cumulative estimations for synthesis of different criteria of decision tree induction. Usage these criteria allow us to produce new type of trees. In this paper we introduce Stable Ordered Fuzzy Decision Tree (FDT). The tree is oriented to parallel and stable processing of input attributes with differing cost. Usage this FDT allows us to realize a sub-optimal classification. Such classification detect a sequence of checks of input attributes with minimize the check-up cost. Also we introduce transformation process from FDT to fuzzy rules set. The results of this paper may be used for design of fuzzy decision-making or expert systems, which based on fuzzy rules set “if x is A and y is B then z is C”