Abstract:

In software reuse, the design model reuse has received more attention. UML class diagram is widely used in software design and the reuse of UML class diagram has become the focus of software design reuse research. As the number of reusable UML class diagrams increases, the classification becomes a fundamental task. A novel two-level classification of UML class diagrams is proposed based on semantics. The semantic similarity measure between UML class diagrams is proposed, a maximum capture algorithm is proposed to obtain center class diagrams to identify the characteristics of each category, and a novel ensemble classifier E-KNN is built based on improved KNN algorithm. The effectiveness of the proposed classifier is verified from classification quality and efficiency.

Key words: UML class diagram, classification, semantic similarity, center class diagram, classifier

摘要：

在软件重用中，软件设计重用受到关注。UML类图被广泛应用于软件设计，UML类图的重用已经成为软件设计重用研究的重点。随着可重用的UML类图数量的不断增加，分类成为一项基础性工作。一个新的基于语义的UML类图的两级分类被提出。定义了类图之间的语义相似性度量，提出了行最大值捕捉算法获取中心类图来标识每个类别特征，基于改进的KNN[（K]邻近）算法构建了一个新的集成分类器E-KNN（ExtensibleKNN）。从分类质量和效率上验证了所构建分类器的有效性。

关键词: UML类图, 分类, 语义相似性, 中心类图, 分类器