Predicting Object-Oriented Software Maintainability Using Projection Pursuit Regression
Lijin Wang, Xinxin Hu, Zhengyuan Ning, KE Wen-hua · 2009
This paper presents ongoing work on using projection pursuit regression model to predict object-oriented software maintainability. The maintainability is measured as the number of changes made to code during a maintenance period by means of object-oriented software metrics. To evaluate the benefits of using PPR over nonlinear modeling techniques, we also build artificial neural network model, and multivariate adaptive regression splines model. The models performance is evaluated and compared using leave-one-out cross-validation with RMSE. The results suggest that PPR can predict more accurately than the other two modeling techniques. The study also provided the useful information on how to constructing software quality model.