Evaluation function guided search for fuzzy set covering

Fuzzy set covering was introduced as an extended counterpart of crisp machine learning methods using a separate-and-conquer approach to concept learning. This approach follows a general-to-specific search through a space of partially ordered conjunctive descriptions. The search path followed depends to a large extent on the evaluation function used. This paper investigates the effect of the evaluation function on the quality of induced rules and the number of conjunctions examined.