Knowledge Discovery by Compensatory Fuzzy Rough Predicates

Compensatory Fuzzy Logic (CFL) are fuzzy logic systems, which satisfy axiomatic properties of bivalent logic and Decision Theory simultaneously. There is a coherence between CFL and other theories like classical logic, t-norm, and t-conorm based fuzzy logic, mathematical statistics, and decision theory. Those properties are the basis of transdisciplinary interpretability in relation to natural language. Hence, CFL has the advantage to model easily the problems by using natural and professional language. The main objective of this paper is to propose a method, inspired by rough sets theory (RST), to approximate decision classes by means of two clusters, defined by logic predicates formed on condition attributes. The importance of the method is mainly that it is a compliment, and not a substitute, of other methods for forecasting, in the sense that its results are more useful in the way of linguistic values than in numerical values.