Selecting software reliability models with a neural network meta classifier
Rafael Caiuta, Aurora Pozo, Leonardo Ramos Emmendörfer, Sílvia Regina Vergílio · 2008
Software reliability is one of the most important quality characteristics for almost all systems. The use of a software reliability model to estimate and predict the system reliability level is fundamental to ensure software quality. However, the selection of an appropriate model for a specific case can be very difficult for project managers. This is because, there are several models that can be used and none has proved to perform well considering different projects and databases. Each model is valid only if its assumptions are satisfied. To aim at the task of choosing the best software reliability model for a dataset, this paper presents a meta-learning approach and describes experimental results from the use of a neural network meta classifier for selection among different kind of reliability models. The obtained results validate the idea and are very promising.