Research And Implementation Of Software Reliability Models Based On Neural Network

With the improvement of software status in weapon equipment,our army has higher requirements for software quality.As an important part of general quality characteristics,software reliability has also attracted more attention.Software reliability evaluation is the key to evaluate the reliability level directly,but it is difficult to put into production.The main problem is that the failure data of software is difficult to collect,and the applicability of a single software reliability model is not strong.To solve problems,this paper proposes a software failure data supplement method,which improves the efficiency;establishes a software reliability model based on neural network,which improves the accuracy of the model;realizes the tool of the model,and promotes the software reliability evaluation into production practice.The paper launches research on the following three aspects:(1)Proposed a supplementary method of software failure data.Firstly,this paper analyzes the current problems of software failure data.By introducing the failure data and use cases of system test and validation test,the failure data is supplemented to alleviate the problem that reliability evaluation cannot be carried out due to the lack of failure data.(2)Proposed a software reliability model based on neural network.By studying the classic software reliability model,choose the appropriate model as the base model,use the neural network algorithm to weight the combination of the output of the base model,and realize the automatic configuration of the weights,which improves the prediction accuracy markedly.This paper applies a variety of neural network algorithms to the evaluation and verification of software reliability models,and compares their prediction accuracy,confirming the effectiveness and superiority of this model.(3)Implemented the tool of the model.In this paper,the proposed software reliability model is implemented,and the theoretical research results are transformed into a usable prototype tool,which promotes the further application of software reliability evaluation in production practice.