In the position paper, we described an experiment to verify a hypothesis of correlation between reusability and a quantitative attribute of a software (volume and complexity). In the experiment, we used a machine learning approach consisting of a machine learning algorithm (C4.5). The results of our experiment show that the selected metrics can predict with a high level of accuracy the classes that can be potentially reusable. The predictive model obtained (volume and complexity) gives a very satisfactory results (89,3% of accuracy).
展开▼
