G. Stiglic and P. Kokol (Slovenia)
Software reliability combined multiple models, mimetic classifiers, elimination of outliers, machine learning, software metrics.
In this paper we present a novel method for induction of comprehensible decision tree for software module reliability prediction. This is done by combining two approaches called outlier elimination and combined multiple models which are also called mimetic classifiers. We test the proposed method on industrial dataset where our experimental results show that it is possible to achieve better accuracy and still maintain a reasonable complexity level of the built decision tree to keep it comprehensible.
Important Links:
Go Back
