Research On Software Evolution Analysis Technique Based On Complex Network

Evolution is an important part of the software life cycle.Mining software repository code information can provide more information for software evolution.The software system is complex,so the characteristic of software system structure is complex too.And in order to meet the demand of new applications,the software system has been updated iterative.The study of software evolution characteristics also arouses people's concern gradually,this paper presents a software network community matching method and an evolution anomaly detection method.This helps us to better understand and analyze the features of software evolution,get the the change information of program in the evolution process.further,this also guide the iterative development of software system.Firstly,this paper proposes a method of building software execution network by using software dynamic execution path.By executing software several times in different system parameters,the information for the invoking relationships of the function nodes is traced.Then the execution results are handled and the model of software execution network is established.Secondly,key nodes is mined by algorithm MTKN(MinTopKNodes).In dynamic execution process of the software system,influential key nodes is mined considering the network software in dynamic execution process between the nodes of the calling sequence and the number of calls.Thirdly,algorithm FBM-Gen(Function Belongingness Matrix Generation)based on key nodes is put forward to divide the software network to soft community.Introducing the theory of community,the concept of function membership matrix is put forward to quantify soft community.we make the key nodes as community center,and use the algorithm FBM-Gen proposed to generate the function dependency matrix.Finally,anomaly detection algorithm SECO-Detection(Software Evolutionary Community Outliers Detection)is proposed based on Community matching.the result can be used to analysis the community evolution in the software evolution process,and find out the anomaly nodes which evolute different from others obviously in the same community.The algorithm proposed in this paper are implemented on the platform of Windows in Java language.Through the experiment on real data set and synthetic data set,we validate the efficiency,accuracy and running stability of the algorithm.