Complex Network Based Service Network Analysis And Its Application In Service Recommendation

Service-Oriented Computing(SOC)is a new computing model after object-oriented and component-oriented.With the continuous develop-ment of SOC,the scale and type of services are growing rapidly,and its application fields also expand,gradually forming a service-centric Internet.For the complex "service network",how to recommend suitable services for users has become a hot issue in the field of SOC research.Service recommendation technology is a popular method,which is very effective in solving the problem of information overload.However,there are still some shortcomings in the existing service recommendation methods:Most of the data used by the existing methods is historical service information and some preference data of users.It is difficult for ordinary users to collecte those data.Some methods rely heavily on service description documents.However,with the emergence of new service,the service documents that people can obtained are not complete,and some services do not even provide corresponding service description documents.In order to solve these problems,the researchers have proposed some new methods.They ad vote solving these problems by introducing the interdisciplinary ideas.With the deepening of the research on architecture and the change of software development mode,the research of service recommendation from the structural point of view has gradually become a novel research idea.The proposal of complex network helps to promote the research of system structure.Complex network theory advocates the analysis and control of the system from a global perspective,not from a partial perspective.The theory provides us with new ideas for better understanding of system architecture.Therefore,it becomes the key issue in this paper that how to understand the topological structure of service network from the structural perspective,and then improve the efficiency of service recommendation.Basing on the interdisciplinary idea and the related principles and methods of complex networks,this paper analyzes the topological structure of service networks by utilizing various indicators of complex networks,then proposes corresponding service recommendation methods based on the results of service network structure analysis.Therefore,the research contents and achievements of this paper are as follows:(1)Complex network based Service network analysisThe structure of the service network has a significant impact on the functionality of the service.According to the various service information we have obtained,this paper has constructed different service networks.Those networks have abstracted the inclusion and the composition relationship between services and services.We learn from the principles and methods of complex networks,and introduce various parameters in complex networks to analyze the structure of service networks intensively.Using the real data from Mashup services and Web APIs on ProgrammableWeb,we have revealed some structural features of service networks(degree distribution?degree centrality?closeness centrality etc.).Some of these features are of great significance for guiding service recommendation.(2)Service recommendation method based on service network structure analysisIn view of the deficiency of the existing service recommendation algorithm,this paper proposes a service recommendation method based on service network structure analysis.This method constructs a service inclusion network,a service Synthesis network by using single-mode projection.And the method introduces a composition network mining method to filter invalid edges in the network;meanwhile,based on some features obtained from the analysis of the service network structure,the method evaluates the popularity and important of service,then we put forward the corresponding service recommendation algorithms for different usage scenarios of services to recommend the composable services which meet the needs of users.We conduct the experiment with the real data of Mashup services and Web API which collected from the ProgrammableWeb,then analyze the experimental results.The experimental results show that the proposed method is feasible and effective for recommending the required services for users.This paper has examined the related problems of service computing from the perspective of complex service network,which provides some new ideas for us to research the problems of service recommendation.