Research On MDA-based Decision Service Composition Modeling In GBDSS

Grid Based Decision Support Systems is an intelligent information system constructed on Grid. It provides highly resource sharing and cooperation abilities to help decision makers solve semi-structured or unstructured deision task. Along with the rapidly growing of economic intergration and social information, the decision task becomes more and more complex so that it is hard for singal atomic decision service to slove. Therefore, it is an urgent research issue to design an effective modeling approach of decision service composition that assembly the massive atomic decision services as a composite decision service for enhancing the abilities of complex tasks handling and heterogeneous resources reusing in GBDSS.The modeling of decision service composition in GBDSS includes four aspects. Firstly, a decision service composition platform should be constructed to support the decision service composition modeling. Secondly, a unified decision service description model should be proposed to assist the decision service modeling. Thirdly, a simple, rapid and effective composition mechanism is needed to optimize the decision service composition. Fouthly, a service matching and selecting algorithm should be provided for improving the composition efficiency. Aim to solve the issues described above, a MDA based decision service composition modeling approach in GBDSS is proposed in this dissertation.Aim to solve the first issue, an architecture of MDA based decision service composition in GBDSS is proposed, which includes MDA, SOA, Grid and Semantic Web. The system structure, lifecycle and operational mechanism of this architecture are proposed as well. This architecture can give a clear whole perspective of MDA based decision service composition in GBDSS.In order to characterize the structure complexity of GBDSS more clearly and make the structure more open and dynamic, a hierarchical Ontology model is proposed.Then more details about the core Ontology of GBDSS is discussed. In oder to realize MDA based decision service composition, a extended UML class diagram for Ontology modeling as well as the mapping rule between the UML class diagram metamodel and the OWL element is proposed.Aim to solve the second issue, two kinds of GBDSS decision service model is proposed, which are atomic decision service model and composite decision service model. The control flow and data flow schema of composite decision service model is analyzed in detail. An approach which uses extensive OWL-S Service Profile Ontology for decision service model storage is proposed according to the characteristic of these models.Aim to solve the third issue, a MDA based model mapping mechanism is adopted. It realizes the construction and transformation among design model, execution model and storage model for decision service composition. At First, an extended UML Activity Diagram based model for composite decision service design is proposed. Then a mapping mechanism is proposed for mapping this PIM to other two PSMs which are extended OWL-S model for composite decision service storage and BPEL4WS model for composite decision service execution. At Last, a traffic optimal path selection case in Grid-based Urban Public Trasport Decision Support System is discussed to demonstrate the use of semantic Web and this mapping mechanism.Aim to solve the forth issue, an interface based service matching approach is proposed. It utilizes concept similarity calculation and reverse derived mechanism to get the preceding matching service aggregation, and utilizes layer based breadth-first algorithm and back-trace based depth-first algorithm to get the spanning tree of the abstact composite decision service aggregation. An improved multi-objective co-evolution genetic algorithm is proposed for realizing the QoS aggregation based physical composite decision service optimization. The experiment results illustrate the availability and rationality of the proposed algorithm.