Research On Discovery And Composition Technology Of Geographical Knowledge Cloud Service

The human society has quickly entered the unprecedented era of big data, and the mode of knowledge acquisition and application is transforming from individual innovation to collective sharing and collaborative innovation. The increasing knowledge discovery algorithms and decision analysis models in global scope only could create values by collaborative running. Geographical knowledge cloud platform is one of the most challenging frontier topics in spatial data mining, geographical decision analysis and knowledge sharing fields, and is the core technology for future development of modern information services, which could provide common technologies and support platform for Chinese information service industry’s rapidly development and upgrading. Aiming to push forward the application of geographical knowledge cloud service, this paper synthesize multiple advanced theories and technologies of distributed spatial data mining, geographical decision, service computing and cloud computing, and carry out systematic study on knowledge cloud service description, publication, discovery, dynamical composition, optimization and distributed execution. The main research contents and achievements from this thesis are:1. The concept and feature of geographical knowledge cloud service are analyzed, and the novel generalized and narrow sense definition of geographical knowledge cloud is proposed. And then a QoS model composes of general index and domain index is devised. Based on the research, a WSDL-extended QoS description method for geographical knowledge cloud service is presented. In addition, after researching on keyword matching, interface matching and semantic matching based service discovery technology, and according to the fact that geographical knowledge services always need to deal with big data, a data format based service matching method is constructed.2. A Hadoop-based geographical knowledge cloud service registry center and a MapReduce-based parallel service discovery framework are presented. The registry center takes advantage of column-oriented storage mode and establishes a dual index structure base on classes and functions, and realizes the distributed storage of knowledge cloud service description data in HBase. The center could reach high extensibility by horizontal expansion, overcome several shortages of traditional registry center such as poor access performance and complicate data synchronization between multiple nodes, and accelerates the service discovery speed.3. The multiple object optimization problem of geographical knowledge cloud service composition is studied. Inspired by basic ant colony algorithm, a data volume aware multiple object optimization algorithm for service composition is devised for the first time, which named as DVA-MOACO. The algorithm utilizes a multi-index service quality evaluation model, and improve the transition probability by considering data transfer cost and QoS simultaneously when ant finding path. To accelerate the process of algorithm, a parallel version named as PDVA-MOACO is described. The algorithm could reach the Pareto near optimal solution rapidly with better QoS performance and lower data transfer cost from numerous candidate solutions.4. A cloud computing platform based distributed geographical knowledge cloud service composition engine is proposed. Aiming to ensure the efficient and reliable running of composition service, the engine provides functions such as process model validation, parsing, distributed execution and exception handling, etc. The paper presents a process model for composition service which composes of visualization and content information. A process validation and parsing algorithm is devised against the model. The architecture of the engine supports the distributed execution in master-slave mode and P2P mode, and has the ability to handle runtime exception.5. A Case Study of the seismic influence field analysis on Fujian province is present to demonstrate the application of geographical knowledge cloud service on GeoKS-Cloud. This case analyzes the theory and process of service decomposition, deployment, modeling and execution, and illustrates the distributed and collaborative service of geographical knowledge cloud services. The result reveals that the surface PGA as outcome of the cloud service almost conform to oval attenuation law and are regulated by site property, and the cloud service have the advantages of more open, extensiable and configurable when compared to ordinary services. It could not only improve the computing efficency, but also enhance the reuse level of data and algorithms, and cut down the cost and difficulty of problem solving.