The design of decision trees in the framework of granular data and their application to software quality models

Abstract

In this study, we discuss the role of fuzzy sets regarded as a comprehensive algorithmic vehicle supporting the design of decision trees. Fuzzy sets help convert continuous attributes into discrete landmarks – fuzzy sets are afterwards exploited as the basic constructs in the optimization of a decision tree. The concept of fuzzy granulation realized via context-based clustering is aimed at the quantization (discretization) process. In contrast to so-called fuzzy decision trees, we enhance the development methodology of binary (Boolean) decision trees rather than generalizing them to the form of fuzzy constructs. Afterwards we solve the problem of quantifying complexity of software systems in the framework of decision trees. The advantages of this approach to quantitative software engineering are discussed in detail. Numerical examples are provided to illustrate the design methodology and provide a better insight into the algorithmic details as well as limitations of the decision trees approach.