Research And Implementation Of Software Reliability Evaluation Technology Based On Model

With the rapid development of information technology,the demand for software in society is gradually increasing.Then,software quality issues are attracting more and more attention.As an important attribute to measure software quality,software reliability is one of the direct user-oriented verification indicators.Software reliability evaluation based on reliability model is one of the most extensive methods to improve software quality,which contains the two most critical processes: the first is parameter estimation,which estimates the parameters of the model to quantify reliability indicators;the second is model prediction,which chooses the appropriate models to obtain more accurate predictions.At present,domestic and foreign scholars still have some limitations in the study of the two processes,which are already the performance bottleneck throughout the reliability evaluation.So,it is necessary to further innovation and exploration.In order to solve the problems of low accuracy,slow convergence,easy to fall into local optimum,and inability to deal with multi-dimensional complex situations in traditional parameter estimation methods,from the perspective of self-adaptation,a parameter estimation algorithm based on optimization ant colony is proposed.By maximum likelihood estimate to get the objective function of the model parameters and the constraint set,and then the definition domain of the parameters is discretized according to the grid method,construct the basic framework of the ant colony search.Finally,strengthening the special nodes improves the convergence speed of the algorithm,and through adaptive adjustment strategy assurance the randomness and diversity of the search.This method overcomes the drawbacks of traditional parameter estimation methods.In addition,although there are more than one hundred kinds of software reliability models and variations that have been published,the simple general model haven't been found.In fact,for the same software,different models will get different reliability evaluation results.With the increasing scale and complexity of the software,the single model can't be fully adapted to the complicated situation.Therefore,this paper proposes a model prediction algorithm based on analytic hierarchy process and neural network.By the model evaluation indicators and the analytic hierarchy process to get the appropriate model,and then the appropriate model is weighted by the neural network to obtain the hybrid model.This method makes different types of models compensate each other,and greatly improves the accuracy and robustness of model prediction.Finally,based on the innovative methods are proposed in this paper,and a model-based software reliability evaluation platform was designed and implemented.The example of the UAV control station system is used to verify that the system is simple to use,the method is effective and has good practical engineering application value.