Title: JAVA APPLICATIONS COMPREHENSION BY DISCOVERING CLUSTERS OF RELATED CLASSES

Year of Publication: 2013
Page Numbers: 281-286
Authors: Jauhar Ali
Conference Name: The Third International Conference on Digital Information Processing and Communications (ICDIPC2013)
- United Arab Emirates

Abstract:


Software maintenance is an important phase of software lifecycle consuming the majority of effort. Program comprehension is the most time consuming activity during software maintenance. Data mining techniques have been used to support program comprehension for large software systems. This paper presents an approach to extract useful knowledge from Java source/byte code, and apply clustering to discover groups of closely related classes. The groups of classes can assist programmers to know the high level structure of large software systems without prior knowledge, and programmers can learn the classes in the same group together. The groups of classes are formed based on different types of coupling between classes, and having common characteristics, such as having the same superclass. A prototype system has been developed and evaluated using a medium sized Java application.