Automated Microservices Identification through Business Process Analysis: A Semantic-driven Clustering Approach

The rise of microservice architectures has further evolved software development practices, building upon the foundation established by web services in breaking down monolithic systems, while offering more finely detailed, coherent, and loosely interconnected services. Despite the transformative potential of microservices, the challenge of effectively pinpointing microservices that meet an organization’s specific requirements remains a difficult task. This paper presents an extended approach for automating microservices identification by leveraging business processes (BPs). Addressing limitations in existing microservices identification techniques, this approach utilizes advanced Natural Language Processing (NLP) techniques—such as Named Entity Recognition (NER) and semantic analysis—to capture dependencies and align boundaries with business logic. This methodology applies unsupervised clustering algorithms, including K-means, K-medoids, and DBSCAN, to generate well-defined microservices by analyzing semantic similarity across BP activities. Evaluation on real-world case studies, such as the Bicing bicycle rental system, Cargo Tracking, and JPetStore, demonstrates its effectiveness, yielding high cohesion, low coupling, and superior granularity in microservices decomposition compared to traditional methods. Metrics such as the Silhouette Index, Afferent Coupling, and Instability further validate its performance. By automating microservices identification with a BP-centered framework, this approach aligns technical architecture with organizational goals, advancing both agility and scalability in software systems. This work marks a step forward in optimizing system modularity and provides a foundation for future scalability improvements.