Abstract

An estimation of change-proneness of parts of asoftware system is an active topic in the area of softwareengineering. Such estimates can be used to predict changesto different classes of a system from one release to the next.They can also be used to estimate and possibly reduce the effortrequired during the development and maintenance phaseby balancing the amount of developers’ time assigned toeach part of a software system. This research work proposesa novel approach to predict changes in an object-orientedsoftware system. The rationale behind this approach is thatin a well-designed software system, feature enhancement orcorrective maintenance should affect a limited amount ofexisting code. Our goal is to quantify this aspect of qualityby assessing the probability that each class will change in afuture generation. Our proposed probabilistic approach usesthe dependencies obtained from the UML diagrams, as wellas other code metrics extracted from source code of severalreleases of a software system using reverse engineeringtechniques. These measures, combined with the change logof the software system and the expected time of next release,are used in an automated manner to predict whether a classwill change in the next release of the software system. Theproposed systematic approach has been evaluated on a multiversionmedium sized open source project namely JFlex,the Fast Scanner Generator for Java. The obtained resultsindicate the simplicity and accuracy of our approach in thecomparison with existing methods referred in the literature.