Comparison research of two typical UML-class-diagram metrics: Experimental software engineering

Measuring UML class diagram complexity can help developers select one with lowest complexity from a variety of different designs with the same functionality; also provide guidance for developing high quality class diagrams. This paper compared the advantages and disadvantages of two typical class-diagram complexity metrics based on statistics and entropy-distance respectively from the view of newly experimental software engineering. 27 class diagrams related to the banking system were classified and predicted their understandability, analyzability and maintainability by means of algorithm C5.0 in well-known software SPSS Clementine. Results showed that UML class diagrams complexity metric based on statistics has higher classification accuracy than that based on entropy-distance.