Research On Software Reliability Model Based On Improved Genetic Neural Network

Software reliability modeling needs in various hypothetical premises, the software of the various productions process data modeling. Hudson software reliability model is established in 1967. Human has published thousands of kinds of model. Even for the same data, there are big differences between the same software system reliability assessments. So the general software reliability model prediction and software reliability model selection are two important aspects of software reliability research. The work of this paper includes two aspects, namely, genetic neural network is adopted to improve the software reliability and software reliability model selection model prediction.Firstly, the introduction of software reliability, software defects, and software, such as the definition of life, introduce classical software reliability models, and compared the classical software reliability models, laid the foundation for the subsequent chapters of theoretical knowledge. Then, in the aspect of software reliability prediction, inspired by Musa execution time model, this paper used an improved genetic neural network model for software reliability. After validation of experimental data, we can see that this improved genetic multilayer forward neural network algorithm and ordinary multi-layer forward neural network reliability model can predict the execution time of the software error. From the ordinary neural network and improved genetic neural network training process of software reliability model, reliability improved genetic multilayer forward neural network model of convergence rate is 2.9 times that of ordinary multi-layer forward neural network. From the ordinary neural network and improved genetic neural network software reliability model error prediction process, improved genetic neural network predicting software reliability model error execution time success rate is 1.64 times that of the normal neural network, achieve the desired research goal.Finally, this paper designs a software reliability modeling software. It mainly based on mixed programming of Matlab and Visual C++. This method not only can quickly analysis and modeling algorithm, but also be able to develop a variety of interactive interface more convenient, and it had good portability and scalability too.