Functional similarity metric for UML models

Abstract

This thesis proposes a functional similarity metric for UML models. In UML, the functionality provided by the software is documented in use-cases which are eventually realized through the interactions of objects. UML models such interactions using sequence diagrams as well as some other similar diagrams. Therefore, we use sequence diagrams in assessing the functional similarity between two software systems. The similarity is assessed by mapping most similar sequence diagrams in the UML models based on matching patterns inside them. Since mapping sets of sequence diagrams was found to be combinatorial optimization problem, heuristic algorithms like Branch and Bound and Genetic Algorithm were used. It was found that Genetic Algorithm performed better especially for large size UML models.