A fuzzy inference framework for induced decision trees

The ID3 algorithm and it's variants have been successfully used to generate decision trees. One predominant weakness of this algorithm is the generation of sharp decision boundaries at every node within the tree. This paper utilises a new Fuzzy Inference Algorithm (FIA) which is used to transform a decision tree into a fuzzy rule set through the process of fuzzification. A Genetic Algorithm (GA) is used to select a series of high performance membership functions which are then applied to branches within a decision tree. The GA will in addition optimise a pre-selected inference technique which will assign a degree of strength to the conjunction and disjunction of membership grades within the tree. FIA has been developed to provide a fuzzy inference framework for the inclusion of different fuzzy inference operators. Five known sets of operators have been applied to two data sets and have shown considerable improvements in classification accuracy over original ID3 (crisp) trees.