Ensemble clustering based approach for software architecture recovery

Frequent software maintenance usually results in a change in the original software architecture and thus architecture recovery is attempted using dependencies in the software artifacts. Many different techniques for architecture recovery are proposed in the literature normally using one out of structural, semantic, and directory dependencies, and applying clustering over these computed values. The main drawback of these approaches is that either they use only one type of dependency computed in a limited way, and or merge multiple dependencies before applying a clustering technique and thus the approaches result in loss of information because each dependency has its unique characteristics. This paper proposes a new approach for architecture recovery using ensemble clustering and utilizing a more precise computation of three types of dependencies: structural, semantic, and directory dependencies. The proposed approach is evaluated over open-source java projects and analyzed. The obtained results clearly show that our proposed architecture recovery approach using ensemble clustering and multiple dependencies performs much better than other conventional recovery approaches. © 2021, Bharati Vidyapeeth's Institute of Computer Applications and Management.