The fuzzy property set model: a fuzzy knowledge representation for inductive learning

Knowledge representation schemes are generally crisp. They allow no expression of the natural concept of the degree to which an object is described by some property (attribute-value pair). We present here a method for representing objects using a fuzzy set of characteristics. The fuzzy property set (FPS) model discussed uses a fuzzy set representation to describe the characteristics of each object in the knowledge system. Thus we can associate a degree between each object and each of its properties. This is an enhancement of the property set model, in which each object is represented by a collection of properties, with no expression of degree. Furthermore, we demonstrate that inductive learning may be performed, using generalized definitions of the rough set upper and lower approximations. The learned concept is represented by its approximations, which in conjunction with a similarity function can be used to rank objects according to their similarity to the concept