Research On Software Reliability Growth Model Based On Fuzzy Wavelet Neural Network

With the rapid development of computer technologies, software products have been widely used in a variety of the fields of society, and software reliability is the increasing important requirement of users. Particularly in the financial industry and aerospace industry, software must have a severely high reliability; otherwise the property and security of people will be under the severe threat and society will suffer a huge damage. Therefore, it is significant to research on a kind of software reliability growth model with generalization and good prediction accuracy.Nowadays, SRGM based on BP Neural Network (BPNN-SRGM) becomes the research hotspot. However, BPNN has some inherent problems: slow convergence, easy in a local minimum value, the limited generalization and difficult to determine the network structure, so these will definitely affect the prediction accuracy of BPNN-SRGM.SRGM based on Fuzzy Wavelet Neural Network (FWNN-SRGM) is presented in the thesis. Firstly, the technology of phase space reconstruction is used to analyze the relation of the failure data. Moreover, training samples are prepared with the result of the phase space reconstruction and the network structure is set up. Finally, BP learning algorithm is employed for training the FWNN-SRGM, and then the trained FWNN-SRGM is used to predict the values of the failure data. The experimental results show that the predict accuracy of the proposed method is better than that of the widely used BPNN-SRGM.Through the above research, the conclusions can be drawn as that the software reliability prediction based on fuzzy wavelet neural network is feasible and better performance can be got by combining the FWNN with phase space reconstruction.