Feature-based critical components identification in multimedia software

Software maintenance is a necessary and frequently occurring activity in software engineering. However, different factors such as inadequate documentation, project size, complex dependencies, and hard to understand architecture cause software maintenance to consume a large part of project resources. Therefore, it is important to assist the newcomers by providing program comprehension facilities that can reveal important information about the software system and can speed up the maintenance tasks. This important information about software includes knowledge about the core part (classes, components, design, etc.) of the system that mainly controls its whole functionality. In literature, different researches attempted to determine core part of the software using various structural, dynamic, and network metrics and termed them as key or critical classes. These approaches have an open scope for modeling coupling relations among different elements of software and most of these approaches need human expertise to identify key classes of the software. Moreover, multimedia software systems are generally interface driven and thus many micro level classes collectively constitute macro level units called as multimedia components. Therefore, this paper focuses to identify key critical units of the multimedia software at component level. The proposed approach in this paper consists of three main phases. In the first phase, different features of a class are identified and assigned a coupling based functional score that represents its significance in the overall functionality of the class. In the second phase, different independent components present in the multimedia software are identified by modeling the system as a dependency graph at the class level. Finally, key critical components of the multimedia software are identified by performing hierarchical agglomerative clustering based on the dependency strength among different identified components. The proposed approach is empirically evaluated on open-source multimedia software of different sizes and the obtained results support the feasibility and usability of the proposed approach of this paper.