Research On Key Technologies Of Network Service Intelligence

The Intelligence of network services has significant roles in promoting the capability of network services and improving user experience. The research on promoting the degree of network service intelligence from the standpoint of machine understanding is not developed. Existing researches reach the aim of machine understanding by using ontology and semantic theory. Some key questions, however, have not been solved completely, such as the discovery of semantic network service, the construction of service model and the semantic interoperation between services. Meanwhile, the intelligence of network services doesn't only contain the network service itself, but also cover the data of network services. The knowledge derived from the data related to network services can provide the valuable information for automatic execution and decision-making. So how to discover the knowledge is one of the key problems for network service intelligence. Based on the previous discussion, the paper will discuss four questions:the discovery of semantic network service, the ontological heterogeneity related to semantic interoperation, the construction of semantic network service model and model transformation, and the knowledge discovery of data related to network services. To fulfill the request of opening network service, the semantic network service adopts the Service-Oriented Architecture (SOA). The discovery of service is the core section of SOA and also the precondition of service execution so the research of service discovery is very essential. In generally, improving the accuracy of service discovery needs to add more elements for discovery, but processing more elements will take more time, which will reduces the efficiency of service discovery. To take account of these two things (accuracy and efficiency) at the same time, the paper proposes an iteration-based heuristic service discovery algorithm. First, the paper divides the matching space into three matching subspaces according to the significance of subspace. The paper finds that the service matching in the previous subspace will affect the service matching in the next subspace so the paper proposes an iteration-based heuristic function. To improve the efficiency of service discovery, the paper imports heuristic information between subspaces. To promoting the accuracy of service discovery, the paper also proposes a weighted ontology similarity algorithm. At last, the experiment demonstrates that the algorithm we proposed doesn't only improve the efficiency of service discovery, but also more accuracy than other algorithms.In the process of semantic interoperation between services, if the ontology is developed by two or more teams, then we may face the problem of ontological heterogeneity so the paper gives an ontology mapping method to solve the previous problem using Bayesian learning. The paper regards the problem that finding the corresponding relationship between the elements of ontology as the problem that searching similar elements. When searching the similar element, the paper doesn't only consider the element itself but also consider the context of the element (the extended set). The paper also uses the String Kernel algorithm to computing the similarity of the name of ontology element in order to improve the accuracy. When the extended set of source ontology element and the one of target ontology element is similar enough, we will change the target ontology element with the source ontology element to reduce the semantic misunderstanding of the ontology due to the difference of element string. Comparing to the existing algorithm, the method we proposed improves the accuracy of ontology mapping.When building the semantic network service, a developer needs to learn the ontology and semantic theory, which may be hard for him/her. Model-driven architecture (MDA) can be used to help the developer to build the semantic network service with UML. But using the MDA to build the semantic network service faces two problems:how to express the service ontology using UML model; and how to transform the UML-based network service model to the OWL/OWL-S based network service. Therefore, the paper uses the ontology definition meta-model (ODM) to solve the problem that how to use UML to express ontology. Then the paper extends the ODM according to the dynamic feature of network services and builds the semantic network service model from two viewpoints-Structure viewpoint and Process viewpoint. More significantly, the paper proposes the automatic model transformation method. The first step is transforming UML model to Profile Ontology using Bayesian learning, and the second step is obtaining the OWL/OWL-S based semantic network service using service discovery according to Profile Ontology. The transformation method we proposed doesn't need to set up transformation rules and can improve the intelligence of model transformation.The knowledge discovery of data related to network services is an essential segment of network service intelligence. For the variety and time sensitivity of data, the paper proposes a method to discovery knowledge among multi-faceted data, which contains time information. First, the paper regards the multi-faceted data as the random variables of Gaussian Random Field, and then discovers the knowledge among multi-faceted data by learning the structure of Gaussian Random Field. The paper calculates the precision matrix of Gaussian Random Field to get the structure of Gaussian Random Field and further get the relationship between multi-faceted data. The method doesn't only get the relationship between data with different types, but also get the relationship between data with different types at different time. The experiment demonstrates that the method we proposed can correctly find the correlation among multi-faceted data.