Structural and semantic similarity measurement of UML sequence diagrams

Software reuse is one of the main best practices in software development process. Software reuse promotes efficiency and effectiveness. Reusing software project artifacts nowadays not only focus on codes but also other preliminary artifacts, such as requirements specification and design models. In order to reuse models, an effort to retrieve similar models from the project repository is a necessary. This paper introduces a similarity measurement on two UML sequence diagrams. The proposed solution uses a greedy approach to calculate the similarity score of two diagrams using four distinctive similarity metrics derived from elements of sequence diagram. This paper also presents the results of measuring the similarity of sequence diagram-pairs in term of accuracy and time.

I. Introduction

Unified Modeling Language is a modeling language standard that has been known and used by software engineers for many years. It plays a main role within software development life cycle of a project [1], [2]. The language allows designers to models the interaction between system and users, interaction between objects, behavior of objects, and implementation and logical structure of the system. These models represent different views and concerns of a single system. Throughout the life cycle, the models may change and evolve due to growing knowledge on the problem domain, lack of knowledge, skills, experience of designers, and constantly changing requirements. Differences between models may also be the result of change-propagation on models of the same software within versions, feature dissimilarities due to specific characteristics of different domains, and other aspects regarding project team attributes, such as experience and skills [3].