Research On Mining Algorithm For Important Nodes From Staric And Dynamic Software Execution Network

With the "Internet +" era of arrival,every walk of life demand for software put forward more and more complex which leads to the larger scale of software system and higher complexity of software structure.Thus,software system in terms of stability,reliability will be under threat.In this paper,we mine important nodes in the software network by studying the structure characteristics and implementation features of software system,which will contribute significantly to locating of bugs and the development and maintenance software.This paper takes the functions of software system as the research object and builds the software execution network of the function layer.According to the spread ability of the bugs to analysis its impact on the entire software system.Thereby defining and quantifying important functions in the software.The main work of this paper is as follows.Firstly,the paper analyzes the relevant network metrics and network characteristics of complex network theory,and puts forward a method of constructing a software execution network by combining with the internal relation of software system.Secondly,From the view of network topology of the software execution network,an important node mining algorithm based on software network topology is proposed.Measuring the difficulty of the spread of the bug by judging each call side for the entire network connectivity.Combined with the network local properties of the node and the global network properties of the edge,NODS and NIDS is defined,and then mining the important nodes in the software system.Thirdly,Fully considering the dynamic implementation feature of the software,an important node mining algorithm based on dynamic implementation of software is proposed.we construct a software dynamic execution network model through a large number of implementation cases.In this model,we define a bug propagation model with the probability of propagation between nodes to search the propagation region ofnodes.Then quantifying of the node's influence accurately by utilizing the call information found in the implementation process,accurate quantification.And then mining the important nodes in the software system.Finally,we validate the results' accuracy of the two algorithms by real data experiments,and analyze the mining performance of the algorithms from of stability,complexity and extensibility.