Research About JDK’s Complex Networks Features

Along with the popularization of the internet and the development of software technology, the size and complexity of the software are increasing constantly. Development software is often in an uncontrolled condition, and wouldn’t guarantee the quality of software products. As a kind of artificial intelligence system, software products’ function, performance and reliability and so on will be affected by the topological structure of the system.In the bottom of software system, the large of accumulation elements, and the relationship between these elements, make software system is difficult to maintain and understanding. Existing software testing method can be said to be a thinking methods: emphasize the relation between the measurable macroscopic quantity which describe the system state measurement and system process. We (Immune-B laboratory) are now going to use another kind of thinking methods to study software systems. To use statistical methods (more precisely, modern complex network method) to study the software system, emphasize the relationship between micro and macro quantitiesDue to the lack of corresponding method, when researchers exam software structure and evolution regularity seldom from the overall and global angle. So we are not very clear to the nature of software. In exploring the structure properties of large software system, complex networks provide a powerful tool support. The research on complex networks is more and more full-blown, that makes that we can apply the research’s result of complex networks to analyze and improve software systems’stability and security.Complex network between the regulation network and random network, is an advanced network. The regulation network and random network are two extremes of simple network, their degree distribution show some "homogeneous". In the actual complex networks, the degree distribution of nodes important is strongly "heterogeneity". In this paper, we only make the relationship of Java JDK’s classes as an instance. Based on it’s statistic characteristic of complex networks, we analyze it’s cluster coefficient, degree distribute and cumulate node degree distribute, and get the trend of Java JDK evolution.We hope that we can from complex networks and complex system angle to examine software. We put the software system as artificial network. From the global and the overall angle exploration and discovery of software system structure characteristics and evolution regularity and thus produce feature. We hope that comprehensive and scientific understandings JDK essential characteristics. To open new ideas for quantize software system’s complexity. Provide a new perspective to software development, software testing, software maintenance and update.This paper finds that:1. As the upgrades of JDK version, its network cluster coefficient and node cluster coefficient increase, the cliques characteristics more perfect.2. Comparison between joining and removing interface, we found that interface contribute to cluster coefficient too much. That is to say:Interface has a pivotal role in making software small collectivization. This conclusion is consistent with the understanding to interface in the software development process.3. No matter what version, Node degree distribution and accumulation degree distribution follow power-law distributions. The distribution of node’s degree of is strongly "heterogeneity". This is typical of the complex network characteristics. Further explanation:JDK class relationship is a complex network.We have seen similarities and differences from these statistical graphics of JDK’s complex network parameters. We hope to find the relationship between these differences among characteristic parameters of complex network and software reliability, safety. We hope that our work to software development and software testing is helpful. I think these are our next work.