An Efficient Scheme for Candidate Solutions of Search-Based Multi-objective Software Remodularization

摘要

Multi-objective search-based software remodularization approaches are used to rearrange the software elements into modules by optimizing several quality criteria. These search-based approaches can find out better quality regrouping solutions compared to the traditional (analytical based) remodularization, if the suitable encoding for the candidate solutions is used. In this paper, we propose an efficient encoding scheme for candidate solutions and use this scheme to remodularize the object-oriented software using genetic based multi-objective evolutionary algorithm. This proposed representation helps in improving human-computer interaction, and semantic based, efficiently-designed and error-free information gets transferred to the computing system through it. To assess the effectiveness of the proposed approach, we evaluate it over six real-world software systems of different characteristics. Further, the approach is compared with the existing encoding scheme (i.e., GNE representation scheme). Experiments show that the proposed approach produces better results in terms of quality, convergence speed and execution time compared to GNE representation scheme.