A clustering-based approach for tracing object-oriented design to requirement

Capturing the traceability relationship between software requirement and design allows developers to check whether the design meets the requirement and to analyze the impact of requirement changes on the design. This paper presents an approach for identifying the classes in object-oriented software design that realizes a given use case, which leverages ideas and technologies from Information Retrieval (IR) and Text Clustering area. First, we represent the use case and all classes as vectors in a vector space constructed with the keywords coming from them. Then, the classes are clustered based on their semantic relevance and the cluster most related to the use case is identified. Finally, we supplement the raw cluster by analyzing structural relationships among classes. We conduct an experiment by using this clustering-based approach to a system- Resource Management Software. We calculate and compare the precision and recall of our approach and nonclustering approaches, and get promising results.