Home // ICSEA 2014, The Ninth International Conference on Software Engineering Advances // View article

Predicting Change Proneness using Object-Oriented Metrics and Machine Learning Algorithms

Authors:
Abdullah Al-Senayen
Abdurhman Al-Sahood
Mohammed Misbhauddin

Keywords: open source software; object-oriented; change proneness; maintainability; prediction

Abstract:
Open Source Software (OSS) has become a huge part of today’s software market and a good source for investments. The establishment of the “National Program for Free & Open Source Software Technology” by the top research center (KACST) in Saudi Arabia to encourage the use of OSS within the community is a major motivation to our work. OSS comes with numerous challenges, one of which is constant change. Being able to identify and measure the change proneness in open source software will ensure saving resources like time and effort. In this paper, we measure the capability of classes of machine learning algorithms to predict change proneness in OSS by using object-oriented metrics. Four classes of machine learning algorithms were considered: Probability-based, Function-based, Instance-based and Tree-based. One complete version of the OSS was used as a training set and tested on the subsequent version to predict the change. The machine learning algorithms were compared based on accuracy, specificity, sensitivity and root mean squared error. We found that nearest neighbor algorithm performed better than the other algorithms in terms of sensitivity and specificity. In the future, we plan to test with different parameters to find a better prediction model for software change proneness.

Pages: 522 to 528

Copyright: Copyright (c) IARIA, 2014

Publication date: October 12, 2014

Published in: conference

ISSN: 2308-4235

ISBN: 978-1-61208-367-4

Location: Nice, France

Dates: from October 12, 2014 to October 16, 2014