Skip to Main Content
Machine learning offers solutions to many real world problems, and many different types of machine learning algorithms were suggested in the literature. Symbolic machine learning algorithms, contrary to numerical type learning like regression and neural networks, try to predict a target concept, but at the same time also explain and justify it's prediction by inducing a rule set. Fuzzy sets are a generalization of crisp sets, and recently much attention was given to fuzzy generalizations of decision trees. Fuzzy rules have also been learned by neural networks and genetic algorithms. However, so far very little has been done to provide fuzzy generalizations of set covering algorithms. In This work we present an algorithm that can induce fuzzy conjunctive rules by following a fuzzy set covering approach. We show the inductive bias of the algorithm is to prefer more general rules with good classification accuracy. We show results on real world domains, and compare the algorithm with other fuzzy and non-fuzzy learning algorithms.