Classification of software artifacts based on structural information

Abstract

Classification of software artifacts, in particularly the source code files, are currently performed by administrator of a repository. Even though there exist automated classification on these repositories, nevertheless existing approach focuses on semantic analysis of keywords found in the artifact. This paper presents the use of structural information, that is the software metrics, in determining the appropriate application domain for a particular artifact. Results obtained from the study show that there is a difference in the metrics' trend between files of different application domain. It is also learned that results obtained using k-nearest neighborhood outperformed C4.5 decision tree and the one generated based on Discriminant Analysis in classifying files of database and graphics domain. © Springer-Verlag 2010.