Predicting modifiability of UML class and sequence diagrams
Matinee Kiewkanya · 2004
Measuring modifiability in design phase can help software designers to decide if the design of the software should be altered in order to improve modifiability of software ultimately implemented. Software metrics is one of effective methods for analyzing software quality. The proposed paper presents a methodology for constructing modifiability model of UML class and sequence diagrams from structural complexity design metrics and aesthetic metrics. Modifiability models are constructed applying three techniques: discriminant analysis, MLP neural network and decision trees in order to find the best one among these models. The obtained models can identify three levels of modifiability of UML class and sequence diagrams: difficult, medium and easy.