Varible Granularity Modeling Of Software Complex Network And Node Importance Measurement Method

Since the computer comes into the world, with the advancement of science and technology, computer hardware develops rapidly. Memory load, the calculation speed of CPU and the speed of hard disk’s reading and writing have increased explosively. With that come to the categories and scale of software application which upgrade continuously. Since computer scientists proposed the software crisis in 1968, the controllability, safety and cost of software development are always the focus of attention.Complex network is a new research topic emerging in recent years, which was applied in the Internet network, neural network, traffic network and other fields widely. The study of complex networks theory applied to the software structure has just started. the existing researches are mainly modeling, measurement, and some achievements of these researches have been obtained. Complex network theory has a broad research space and important practical application value in terms of improving the quality of software by researching software structure.This paper first presents a variable size to describe complex network modeling method of software architecture. The methods make use of the call existing in software, adopt the algorithm of priority to the depth to traverse each method in the software source code, which establishes a node for each of the variables. And each set of variables effect relationship establishes an edge. After the traversal puts to an end, it also builds a variable granularity software complex network model, calculates the metric value of complex network in the model and analyzes whether the characteristics of small world and scale-free of the model are obvious.The paper then presents a node metrics method named NIESN which is used in complex network describing the software structure. And this method takes into account the specific characteristics of complex network describing the software structure description of complex network software structure. When an edge exists between two,the two vote for each other through recursive algorithm. Finally, the final results of measurement will be obtained when the metric values for all the tend to be stable.This paper finally carries on two experiments on a large number of real software. One is the modeling, and the other measures the importance of. The results show that the model proposed in this paper has good practical application values. Compared to previous methods, the node metrics method is more suitable for the measurement of node importance in complex network describing software structure.