This paper introduces an approximate reasoning system for assessing software quality and introduces the application of two computational intelligence methods in designing a software quality decision system, namely, granulation from fuzzy sets and rule-derivation from rough sets. This research is part of a computational intelligent systems approach to software quality evaluation, which includes a fuzzy-neural software quality factor-criteria selection model with learning and a rough-fuzzy-neural software quality decision system. Overall, computational intelligence results from a synergy of various combinations of genetic, fuzzy, rough and neural computing in designing engineering systems. Based on observations concerning software quality and the granulations of measurements in an extended form of the McCall software quality measurement framework, an approach to deriving rules about software quality is given. Quality decision rules express relationships between evaluations of software quality criteria measurements. A quality decision table is constructed relative to the degree of membership of each software quality measurement in particular granules. Decision-tables themselves are as a collection of sensors, which "sense" inputs and output conditions for rules. Rosetta is used to generate quality decision rules. The approach described in this paper illustrates the combined application of fuzzy sets and rough sets in developing a software quality decision system.
Discussion(0)
No comments yet. Be the first to comment.