Interactive scalable condensation of reverse engineered UML class diagrams for software comprehension

Osman, M.H.B. (2015)

Doctoral Thesis

Software design documentation is a valuable aid in software comprehension. However, keeping the software design up-to-date with evolving source code is challenging and time-consuming. Reverse engineering is one of the options for recovering software architecture from the implementation code. However, reverse engineering techniques suffer from several problems; one of them is that the resultant diagrams offer too detailed information. Therefore, this thesis aims to provide an automated framework to simplify reverse engineered diagrams (specifically class diagram) for assisting software comprehension. This thesis presents a study to elicit software developer’s views on simplifying class diagrams. It also explores the suitability of object-oriented metrics as features for the class diagram simplification. This thesis introduces a novel approach for classifying class inclusion/exclusion by using object-oriented design metrics as features. In addition, a set of text metrics based on class names is invented to enhance the performance prediction by combining these text-based features with object-oriented design metrics. Based on this approach, an automated tool is developed to support software comprehension by providing interactive explorations of various design abstraction levels. Finally, the proposed framework and tool are validated through a user study.

• Reverse-Engineering
• Supervised Machine Learning
• Class Diagram Simplification
• Unified Modeling Language

• All authors: Osman, M.H.B.
• Supervisor: Kok, J.N.; Chaudron, M.R.V.
• Co-supervisor: Putten, P.W.H. van der
• Committee: Bäck, T.H.W.; Guéhéneuc, Y.G.; Vinju, J.J.; Verbeek, F.
• Qualification: Doctor (dr.)
• Awarding Institution: Leiden Institute of Advanced Computer Science (LIACS), Faculty of Science, Leiden University
• Date: 2015-03-10
• ISBN (print): 9789462595880