Research of Software Faults Prediction Model Based on Artificial Neural Network
Qi Jie Wang · Jisuanji fangzhen · 2005
Predicting the fault-prone software module early in the software development can improve the software quality significantly. A common problem in software faults prediction is the presence of noise in the data. Neural Networks are robust and have a good noise tolerance. This paper presents a kind of software faults prediction model based on artificial neural network and the structure of the feed-forward multi-layer network with backpropagation learning algorithm. We use this approach to analyse the SDH telecommunication software which was developed by Lucent Technology Optical Network Company and get high predicting accuracy. We select training sets and test sets in many different software releases and discuss the relation between training set and the prediction accuracy.