On The Use Of KStar Algorithm For Predicting Object-Oriented Software Maintainability

Abstract

This paper presents an ongoing work on using KStar algorithm to predict Object-Oriented software maintainability. The maintainability is measured as the number of changes made to code throughout the maintenance period by using Object-Oriented software metrics. We build a prediction model based on data collected from two different Object-Oriented systems. However, to figure out the advantages of KStar algorithm we made five experiments with the Weka machine learning workbench and to compare our proposed model with the other algorithms which are (linear Regression, Neural Network, Decision Tree, SVM), the prediction accuracy of all models is evaluated and compared using cross-validation and different types of accuracy measures. As a result, KStar yields better results and it demonstrated to be the best of them to predict more accurately than the other typical techniques.