Ranking Algorithm Of Class Influence In Complex Software System

Studies show that most software defects are contained in fewer module,the spread of these defects between modules become important issues affecting the quality of software.At present the studies of defect spreading influence are more applied in the smaller software system than in large-scale software system and there are less studies on defect propagation effects in the large-scale software systems as well as identifying important module.According to the above problem,we will extract software networks from codes and investigate the spread of module defeats in software system through ranking the influence of modules.In this paper,we summarize the important node identification algorithms in recent years and propose a new algorithm based on LeaderRank,named SWNodeRank,which is used to identify important nodes according to the nodes' internal features and external connectivity.It is to make original graph strong connected by adding a virtual node and virtual edges.Nodes assign importance score according to the weights on edges.This process is iterative until the scores become stable and finally recognize strong spreading important nodes.In this paper,we evaluate the accuracy of the SWNoderRank and spreading range through the use of Account,Junit and Jedit as data set,compared with LeaderRank and weighted LeaderRank.The experiment has verified two aspects:first,SWNodeRank could identify influential spreader more effectively than LeaderRank and weighted LeaderRank;second,compared with LeaderRank and weighted LeaderRank,SWNodeRank is less affected with different infected possibility threshold.