Recovery of Architecture Module Views using an Optimized Algorithm Based on Design Structure Matrices

TL;DR

This paper demonstrates that optimized clustering of design structure matrices can effectively recover software architecture module views, offering a viable alternative to existing methods.

Contribution

It adapts and optimizes an existing DSM clustering algorithm for software architecture recovery, demonstrating its effectiveness in this domain.

Findings

DSM clustering is a viable alternative for architecture view recovery

Optimization through parameter tuning improves clustering performance

Results confirm the method's applicability to software architecture analysis

Abstract

Design structure matrices (DSMs) are useful to represent high-level system structure, modeling interactions between design entities. DSMs are used for many visualization and abstraction activities. In this work, we propose the use of an existing DSM clustering algorithm to recover software architecture module views. To make it suitable to this domain, optimization has proved necessary. It was achieved through performance analysis and parameter tuning on the original algorithm. Results show that DSM clustering can be an alternative to other clustering algorithms.