Software Reliability Prediction via Neural Network
Wataru Zama, Xiao Xiao · 2023
Currently, most systems are composed of software, but defects caused by software faults may occur, causing various problems. Therefore, by analyzing the number of faults detected per unit time in the testing phase of the software development process and predicting the number of faults that will be detected in the future, the reliability of software can be quantitatively evaluated. This allows developers to accurately assess the current and future reliability of their products, estimate the man-months required for testing, and predict appropriate shipping times. In this paper, we combine Recurrent Neural Network (RNN), which is one of the neural networks, and Wavelet Shrinkage Estimation (WSE), which is a method for analyzing and estimating the number of software faults, for the prediction. We conduct experiments based on real data analysis and discuss the effectiveness of our proposal.