Fuzzy label semantics for data mining

This chapter gives a tutorial introduction on the label semantics framework for reasoning with uncertainty and several data mining models which are developed based on this framework. Modelling real world problems typically involves processing uncertainty of two distinct types. These are uncertainty arising from a lack of knowledge relating to concepts which, in the sense of classical logic, may be well defined and uncertainty due to inherent vagueness in concepts themselves. Traditionally, these two types of uncertainties are modeled in terms of probability theory and fuzzy set theory, respectively. Zadeh Zadeh-IS2005 recently argued that all the approaches for uncertainty modelling can be unified into a general theory of uncertainty (GTU). In this chapter, we will introduce an alternate approach for modelling uncertainties by using random set and fuzzy logic. This framework is referred to as label semantics where the labels could be discrete or fuzzy labels. Based on this framework, we proposed several new data mining models. These models not only give comparable accuracy to other well-known data mining models, but also high transparency by which we understand how classifications or predictions have been made instead of a black box.