| Since 21st century, the rapid developing of information technology and computer science and the expansion of its application lead to the increasing scalability and complexity of software system, which constantly generate problems related to software security and hinder the effective application for security, reliability and controllability of the software. To turn the situation around, the new ideas about software technology and methods are needed.Software structure has impact on the function, capability, security and reliability of software. The focus of the past researches was almost emphasizing in the locality of the software system while ignoring the global properties. Therefore, it leads to lack clear comprehension of the essence characteristic of the software. The complex network theory provides powerful support to the research of the global structure of the software for the reason that it ignores the details and emphasizes the universal objective principles. Thus, the researches about the unique characteristic and evolving model of the topology of the software structure have aroused many people's interest.In the structured software, the essential principle followed by the high-quality software is low coupling. Coupling is the most important inner characteristic. Coupling is symbolized as the characteristic of interaction of elements. It directly affects the complexity of the interaction of the elements, which is a very important index to judge the complexity of the software structure. Therefore, it becomes a key research that the relationship between coupling characteristic and the structure and the impact to the structure characteristics introduced by coupling.There are several innovative contributions in this paper and its major work is:(1) Empirical analysis of software coupling networks is processed. In addition, the related conclusions are gotten and some existing ones are amended.By empirical analysis of software coupling networks, the conclusion that the degree distribution obeys power-law is amended. At the same time, the degree distribution of the software coupling networks which has the phenomenon of stability of exponent is first found and this is very helpful to understand further the inner structure of the software for the designers of software system.(2) The multi-coupling software networks clusters are established and the topology of sub-networks with different types of coupling are respectively analyzed, and then the correlation of coupling characteristic and topology characteristic is found.Based on the principle of decomposition of networks, the software coupling network is decomposed into several different sub-networks depending on their types of coupling and multi-coupling software networks clusters are established. After the comparison and analysis, the same rule is found in their individual sub-networks on the characteristic of coupling, such as the perfect stability of in-degree distribution in contrast to huge variation of out-degree distribution. Further, the reasons producing the phenomenon from the angle of inside formation mechanism are analyzed.(3) Weighted software network based on the coupling frequency is established and the correlation between the distribution of weight (coupling frequency) and the software structure is found.Based on the methods and theory of weighted network, the weight is added to the edges in software coupling network after taking accounting into coupling frequency between classes. Then the weighted network based on the coupling frequency is established. Through studying the topology of weighted network of software, we find weight (coupling frequency) has the same distribution as degree. In addition, the exact functional relation of the two is found. The models provide more effective method to overall comprehend the relationship of the coupling characteristic and software structure.(4) The measuring parameters that depict the coupling characteristic of class level and cohesion characteristic of core functional modules are proposed.The researches on the property of complex network in the traditional object-oriented coupling measurement are conducted, and the reasons are analyzed from the principles and mechanism of software designing. Further, k-core decomposition is applied to study the hierarchical structure of the coupling software networks and the coupling characteristic of class level and cohesion characteristic of the core-function modules are analyzed. Moreover, measurements of coupling in the class level and of cohesion in the core function are proposed. They provide effective solution to mastering the troubles caused by the two characteristics.(5) The dynamic evolving model of software coupling networks with multi-links is established.Through considering the coupling characteristic between classes and the dynamic evolution of software coupling network, the dynamic evolving model of software coupling networks with multi-links is first built. After that, the analysis of dynamic stability of the model is conducted. At the same time, the sufficient condition for the stability of the dynamic model is derived. At last, some simulations are implemented and it well reflects the scale-free characteristic of the software system according to the practical software system data of the topology. The model reveals some connections between coupling of classes and complexity of software system. It is very instructive to understand the impact of the inner characteristics to the quality of software systems for the designers. |
PDF Full Text Request