A systematic approach for layered component identification

Understanding a system from architecture view is very important for varieties of reasons. After frequent modifications, the system documents become outdated since they are inconsistent with the behaviors of the system. Currently, many component identification processes have been provided - clustering analysis, matrix analysis, graph-based decomposing methods, etc. - to understand the system design from source code. Also many metrics have been defined for the distance of software entities. However, the identification results by these approaches still face lack of proper organization and understandability. This paper proposes a systematic process for layered components identification by combining business process analysis and software clustering techniques. The approach utilizes the business rules of the system and greatly improves the comprehensibility and maintainability of the target component-based system by assigning the function description and structure info to the components.