Research On Reliability Growth Model Of Imperfect Debugging Software Based On Neural Network

The performance of software reliability evaluation directly affects the workload of software testing,and it is a very challenging task to accurately predict software reliability.In recent years,due to the development and popularization of computer technology,the functions and uses of computer software are also complex and diverse,and because it plays an increasingly important role in human society,people's production and life has become increasingly inseparable from the participation of computer software.Therefore,more and more attention has been paid to the quality of software.Software reliability is an important attribute to measure the quality of software.How to release high-quality software products is a very difficult thing,so the research on software reliability as the core has always been a hot topic.Software reliability growth models(SRGMs)is an important mathematical tool to measure software reliability,which plays an important role in software reliability evaluation,assurance,test resource control and optimal release time research.So far,hundreds of software reliability growth models have been developed for practical projects.These developed models are based on a variety of assumptions,so the universality and prediction accuracy of the model is not high.Because ANN(artificial neural networks)has a strong non-linear computing ability,the software reliability growth model based on ANN can obtain higher prediction accuracy and stronger universality.Due to the strong coupling of modern software integration scale and the powerful functions of software system,when the detected faults are removed,it is likely to introduce new faults.When it is impossible to avoid the impact of the number of introduced faults on the software testing process,it is of great significance to study the process of fault introduction on the intention of software reliability.Because of the delay of software system,we know that when the software fault is detected,it cannot be eliminated immediately.There is a time delay between fault detection and troubleshooting.When the impact of time delay cannot be ignored,the delay time becomes an important factor affecting the reliability of software.The software reliability growth model based on neural network not only inherits all the attributes of traditional SRGM,but also has strong nonlinear operation ability,distributed storage and learning ability.In this paper,the mapping relationship between traditional SRGM and neural network in software testing is studied deeply,and a software reliability growth model considering fault introduction and delay in troubleshooting is established.On this basis,the traditional model and dynamic weighted combination model considering fault introduction are discussed,as well as the traditional model and delay in troubleshooting The advantages and disadvantages of the delayed generalized dynamic integrated neural network model.The main research work of this paper includes the following aspects:(1)This paper first summarizes the traditional software reliability growth model,and finds out the mapping rules between the traditional SRGM and the neural network model through the example of the G-O model.At last,the modeling process of each neural network model is explained in detail,and it is verified on the failure data set.(2)In the process of software testing,when the detected faults are eliminated,new faults may be introduced.Therefore,this paper establishes a software reliability growth model of dynamic weighted composite neural network considering fault introduction phenomenon,and verifies it with corresponding fault data set.The experimental results show that the software reliability growth model of dynamic weighted composite neural network considering fault introduction phenomenon has better fitting effect and prediction performance.(3)In the process of software testing,there is a delay time between fault detection and troubleshooting.When the fault is detected,it cannot be eliminated immediately.Therefore,this paper establishes a general dynamic integration software reliability growth model considering the waiting delay of troubleshooting,and verifies it with the corresponding fault data set.The experimental results show that the general dynamic integration neural network software reliability model considering the waiting delay of troubleshooting has better fitting effect and prediction performance.