Prediction Algorithms and Consecutive Estimation of Software Product Line Feature Model Usability

Abstract

Evaluating the quality of software product line (SPL) feature models is essential because a low quality design will be eventually reflected in almost all the products of the family. Assessment of usability in particular improves reusability which is the ultimate aim of SPL. Because feature models are used in the early stages of development, their usability assessment will help developers to design highly useable product lines. In this paper, it is proposed to develop a prediction model which can be used to forecast the usability of feature models. In our previous work we had proposed metrics for usability assessment of feature models. In this current work, we have used these metrics to develop and compare prediction accuracy of five machine learning algorithms in terms of usability and predict usability of feature models on the basis of the best performing model.