Software Remodularization for Maintainability and Evolution: A Systematic Mapping Study of Techniques, Tools, and Open Challenges

The remodularization of software has become one of the most important activities in software maintenance and evolution, and the presence of the multi-objective problem of continually degrading the quality of software architecture in response to new requirements. This paper conducts a systematic mapping study of remodularization research published between 2010 and 2025, with the aim of classifying the techniques, tools, evaluation practices, and identifying open challenges. Our systematic literature analysis aims at building a broad classification, quantifies trends in methods and evaluations, and uncovering gaps in industrial adoption and modern AI/ML-based approaches. changes has been met by the wide utilization of clustering and search-based methods, whereby the traditional clustering tools and algorithms (e.g., Bunch and variants of spectral/hierarchical methods) represent a large fraction of the literature, and the search-based/metaheuristic models (GA/NSGA variants and hill-climbing) Semantically signalled information retrieval and topic-modeling algorithms (LSI/LDA/RTM) have been applied to modularization and refactoring recommendation and more recent deep learning / big-code methods (code embeddings, GNNs) are being investigated but are under-represented in empirical validation. Assessment is mostly based on structural measures (MQ, coupling, cohesion) and open-source system case studies, and few industrial replications and limited longitudinal research of long-term maintainability benefits are instantiated. The mapping identifies such gaps in the maturity of tools, explainability to the ML methods, and remodularization to the cloud-native/microservice domain.