Research On The Open Source Software Reliability Analysis And Evaluation Based On The Change-point

With advantage of short development cycle,many version information and fast update speed,open source software systems are widely used and their reliability gets more and more attention.Compared to the traditional closed source software,the debugging of the open source software is largely depending on the community members and users.Some abrupt changes of the fault detection are likely to happen nearby the release time.Therefore,aiming at the error characteristics of open source software,this paper consi ders the change point problem on the basis of the existing software reliability growth model to achieve more accurate model fitting and data prediction.At present,in the research of software reliability model,the chagne point problem is mainly divided into two kinds of methods: parametric type and nonparametric type(this paper,the nonparametric methods of the control chart,arithmetic mean,release time)two methods are studied.In the nonparametric method,the method of wavelet analysis is added.Based on the failure data of specific open source software,the characteristics of parametric and nonparametric methods are studied in this paper.By considering different types of software reliability growth model,tested on the different open source software datasets,combined with the optimal model selection strategy,this paper proposes an open source software change point selection method.At the same time,according to the character istics of different failure data,two kinds of mixed models are proposed,which can be subject to different distribution and failure detection process.It is found that when there is a proper need in predicting the model obtained by the two models has better effect.