Software Classification Using Structure-Based Descriptors

• Main Author: Ramadan, Qusai Hussein
• Format: Thesis
• Language: English; English
• Published: 2009
• Subjects: QA76 Computer software
• Online Access: https://etd.uum.edu.my/2043/1/Qusai_Hussein_Ramadan.pdf; https://etd.uum.edu.my/2043/2/1.Qusai_Hussein_Ramadan.pdf; https://etd.uum.edu.my/2043/

With the huge increase of software functionalities, sizes and application domain, the difficulty of categorizing and classifying software packages for reuse and maintenance purposes is on demand. Building automatic classification mechanism will help to save the budget, time, and the efforts of the organizations, especially the administrators of software repositories. This work includes the use of structure information contained in source code programs to automate program classification. Three software metrics namely; LOC, MVG and WMCl have been extracted from programs of category board and puzzle obtained from SourceForge.net. A total of 2800 programs have been used during the training process while two different datasets of size (28) were used for testing. Based on the undertaken experiment, the IBK algorithm is noted to generate the highest classification accuracy (74.8%) compared to several other algorithms provided in the Weka tool. The study also shows that board programs are written in different structure compared to the puzzle programs. Hence, showing that structure information can be used to classify programs into application domain.