Software fault prediction using neuro-fuzzy network and evolutionary learning approach

In the real world, a great deal of information is provided by human experts that normally do not conform to the rules of physics, but describe the complicated systems by a set of incomplete or vague statements. The need of conducting uncertainty analysis in software reliability for the large and complex system is demanding. For large complex systems made up of many components, the uncertainty of each individual parameter amplifies the uncertainty of the total system reliability. In this paper, to overcome with the problem of uncertainty in software development process and environment, a neuro-fuzzy modeling has been proposed for software fault prediction. The training of the proposed neuro-fuzzy model has been done with genetic algorithm and back-propagation learning algorithm. The proposed model has been validated using some real software failure data. The efficiency of the two learning algorithms has been compared with various fuzzy and statistical time series-based forecasting algorithms on the basis of their prediction ability.