Machine Learning-Driven Service Boundary Detection for Effective Legacy System Modernization

Abstract

The evolution of legacy software systems presents an ongoing challenge for software engineering due to aging architectures, lack of documentation, and the increasing need for modular, maintainable, and scalable systems. This research examines the integration of machine learning-assisted techniques in the modularization of legacy systems, particularly through service boundary detection, as a means to support modernization efforts. By analyzing the convergence of traditional software refactoring approaches, automated service identification, and contemporary frameworks for legacy system assessment, this study situates machine learning as a transformative tool within legacy software evolution. The investigation begins with a theoretical foundation of legacy system characteristics and the historical progression of modernization strategies, emphasizing the limitations of manual refactoring and the benefits of intelligent automation. Methodologically, the study employs a detailed qualitative and conceptual framework derived from empirical literature, including case studies, expert surveys, and comparative analyses of software modularization techniques. The results are interpreted descriptively, highlighting the ability of machine learning models to identify latent service boundaries, reduce coupling, and improve system maintainability while ensuring functional integrity. The discussion synthesizes these findings with broader theoretical debates on software engineering practices, including cost-benefit considerations, risk management, and organizational adoption of AI-assisted techniques. Limitations are acknowledged concerning model generalizability and domain specificity, and recommendations for future research include cross-domain validation, integration with service-oriented architecture migration, and the development of explainable machine learning models for practitioner trust. This comprehensive study establishes a foundation for integrating artificial intelligence into legacy system modernization, positioning machine learning as a critical enabler for sustainable software evolution.