Fuzzy clustering of software metrics

We investigate the use of fuzzy clustering for the analysis of software metrics databases. Software metrics are collected at various points during software development, in order to monitor and control the quality of a software product. We use fuzzy clustering to examine three collections of software metrics. This is one of the very few attempts to use unsupervised learning in the software metrics domain, even though unsupervised learning seems more appropriate for this application domain. Some characteristics of this application domain that have significant implications for machine learning are highlighted and discussed. Our results illustrate how unsupervised learning can be used in software quality control.