| Software is the effective tool for dealing with complex problems nowadays, with the problem domain of deepening and expanding, the complexity and scale of software grow with each passing day, Developers are increasingly difficult to understand the development of software system, The risk of development enlarges gradually, and software is different from the general industry product whose quality can be constrained by strict testing standards, It is difficult to get effective control and guarantee. The traditional software engineering theory showed great limitations in large-scale software development, such as evaluation, maintenance, etc, it is approaching the extreme limit of its complexity. Therefore, How to understand and quantify the software complexity, Construct the good architecture of adapting to maintain, rational evaluate software system, It is not only the key to software metrics, But also is a great challenge for software engineering.The software system is a kind of complex system that composed of a large number of bottom elements and their interactive relationship, its structure contains all the software topology information and networking features, Software structure determines its function and quality. The traditional software engineering design methods generally only concerns software underlying local structural features, and neglects the more important global structure of reflect the overall characteristics, the developer is difficult to grasp the scale of software system design, use and research, leading them to the lack of clear macroscopic understanding in the software engineering science and basic structure characteristic, which in turn prevents them from efficient understanding, reusing the existing system, and getting high quality development. In recent years, the researchers found that, the internal structure of software shows obvious characteristics of complex network, So using complex systems and software engineering, Making interdisciplinary research, from the complex network and complex systems perspective to re-examine the software system, it can be abstracted into artificial complex network to explore software structure complexity essence, evolution and behavior characteristics and the quantitative evaluation, which forms a new network concept to guide the large-scale software system design, break through the traditional way of thinking, control and improve software quality from the source, ensure the software performance, reduce maintenance costs.Firstly, this paper analyzes the traditional measurement methods deficiency in large-scale software application, evaluates software system from the new perspective of complex network. Focus on the analysis and measurement of global software structure characteristics, develops an analysis tool to realize topological analysis and visualization, Builds the software static structure network model, Intuitive and cursory evaluates software system; Builds the software sample database for follow-up research; After a large sample, multi dimension software structure characteristics and correlation of statistical analysis, elaborated it effect in software structure complexity, stability and reliability, Combined with the new definition of dependence, the influence and entropy concept of thermodynamic, from different perspectives to construct two quantitative description of system complexity, preliminary determine reference measurement parameters of both node and structure dimensions.Subsequently, this paper from the base software structure, defines the software network core concept, based on the nuclear structure extraction and extended of software, using coreness and k-nuclear decomposition analyzed the potential structure characteristics of large-scale software, combined with the idea of software design, quantitative analyzed the important position and effect of coreness in software hierarchy, centrality, connection trend and evolutionary trend, Formed the initial and reasonable software design criterion, provided guidance and advice for simplifying the understanding of large-scale software systems, targeted testing and measurement.Secondly, in comparison to the conventional measurement method, Presents a 2D software measurement system based on complex network (BCN measure and evaluation system). By large samples summarized function and use of some characteristics in software analysis such as the degree distribution, clustering coefficient, average shortest path, Betweenness, structure entropy, coreness, according to the statistical analysis results for a large of the best open source software, determined the good structure design criterion of software from statistics as a quality assessment reference; From the microscopic and macroscopic topology dimensions to define eight metrics to measure the important unit and structure characteristics of software system globally, quantify the software quality evaluation, and through the actual hydropower simulation system to verify the metrics system is reasonable, feasible and effective.Again, based on the measurement methodology studied the software system evolution laws from time dimension, focused on the software network average node degree, average number of clusters, Betweenness, coreness, degree correlations, matching coefficient and other important features with large sample version change tracking statistics analysis, studied its change characteristic, revealed software evolution model, forms and basic rule, provided reliable basis for understanding the ecological features of software, software evolution trend prediction, and guiding software project development.Finally, through the standard structure entropy study in the software ecosystem, explained the order in software network evolution, further explained the software evolution direction is enhanced order, incremental complexity and steady topological structure; revealed the software system is a dissipative structure, evolution make the software has significant ecological characteristics,metabolic, organized, orderly and mutations, the final evolution trend is obtained by adaptive stabilization; defined evolutionary rate and gave the quantitative discriminate method of evaluative speed by grey theory; proposed a new prediction method for software complexity evolution using logistic equation, and provided reference for the software development and quality control. |
PDF Full Text Request