Research On Key Function Mining Method In Dynamic Software Structure

With the development of Internet technology,software products play a more and more important role in people's daily life,and the function of software is also enriched,which leads to the complexity of software system structure,also makes software development more and more difficult.The researchers found that in the software structure,function is the smallest unit causing errors and defects,therefore,analysis of the key node in software system structure is very important for maintaining software system structure,improving the speed of software development and ensuring software quality.This paper analyzes and discusses the dynamic software structure,takes the function as the basic research granularity,and designs two different algorithmswhichc can mine two kind of key nodes by analyzing the behavior properties of the function in the software execution process,and analyzes the advantages and accuracy of the algorithms.First,in the Linux operating system,GCC compiler tool,tracking tool Pvtrace and visual tool Graphviz are used to extract and visualize the software execution process,thus the function execution sequence and function call graph in the software running process are obtained,which are mapped to the sequence model and the complex network model respectively.Secondly,according to the relevant information of function execution frequency in the sequence model,the frequent key node mining algorithm DNFM is designed.Using the function execution order,the call relationship and other information in the model,the node importance is evaluated from the point of function execution frequency,and preventing the occurance of errors of the function which were executed frequently,ensuring the quality of the software.Thirdly,according to the related information about function correlation in the sequence model and the complex network model,the association node mining algorithm PNMEA is designed to excavate the key nodes which are closely related to other functional entities,and provide some theoretical guidance and technical support for the location and prevention of errors of parameter transmission.Finally,the two algorithms are implemented to verify the effectiveness.At the same time,this algorithms are compared with classical algorithms including PageRank,Betweenness Centrality,Closeness Centrality,Degree Centrality and infectious disease transmission model SI,in order to prove that the algorithms can solve the disadvantages of the classical algorithms that can not analyze the features of the software structure,in addition,the algorithms can find key nodes of different types in the dynamic software structure.