Source code dependency graph based contextual probabilistic clustering approach on large open source projects

Abstract

Software systems tend to become larger and more complex in terms of different metrics such as fields, methods and classes as the functional requirements of the project increases. Most of the open source or commercial software projects are represented in simple diagrammatic representation in order to understand the analysis of the source code metrics. Also, these projects are difficult to analyze the source code metrics by using traditional similarity measures and graph dependency approaches due to high computational memory and time. Analysis of source code metrics and compiled class metrics are necessary to represent various metrics and its complex relationship for software design representation. In order to overcome these issues, a novel graph dependency based probabilistic similarity clustering approach is implemented on the source codes and class files metrics. Experimental results proved that the present approach is better than the traditional source code analysis methods in terms of contextual similarity and accuracy.