Study On Macro-economic Intelligent Forecasting Decision Support System Based On Agent

The macro-economic decision is the striking point of macro-economic management, the decision impacts of the efficiency and benefit of national economic. The macro-economic forecasting and decision's comprehensive integration system has become a research emphasis in the field of economic forecasting, and intelligent comprehensive integration system has become an important tendency of the development of this field. The paper encloses the task of project of National Natural Science Fund, a multiobjective and dynamic input-output optimization model, algorithm and application, and conbines models and methods of macro-economic and IDSS together. The achievement can be summarized as follows:1. The present research of macro-economic forecasting IDSS and Agent technology are studied. In view of the problems and deficiencies of FDSS existed, for example, the interactive interface, management and gain of knowledge, forecasting models etc., the paper proposes to structure and realizes MEIFDSS using the new AI technology - Agent, and paper also shows the whole structure of the system, which will take advantage of the virtue of Agent-based systems.2. A task-oriented Agent design method is proposed, it designs the structure of Agent based on the tasks of Agent. Under the general framework, Agent's structures are very different because of their different specific tasks. Considering the need of system in actual, the paper designs the scheme that seven different kinds of Agents to finish tasks all together, including interface Agent, center control Agent, model Agent, analysis Agent, integration Agent, information Agent and model building Agent, and the different type Agents' concrete structures are appeared.3. The mapping relation from math model to Agent is proposed. A mode, which uses Agent technology to structure the model system, is given. It treats the applying conditions, advantages, shortcoming and the out variables of model as theknowledge of Agent, which makes the model system selfadaptation and initiative in the environment. It presents a new thoughts and method for the encapsulation of models to agents.4. In multi- agent system, the cooperation among Agents is a most important subject for study. The paper also proposes to build a special Cooperation Agent in the system in order to complete the cooperation among Agents in the multi-Agent system. All of the protocol, policy, disposal and evaluation of cooperation are done by cooperation agent. It will make other Agent structure simple, let cooperation more flexible, operation easier.5. An experimental system of MEIFDSS in condition of single computer is designed and realized. It has forecasting and simple analysis decision functions. It contains multi-objective dynamic input-output optimization models and artificial neural networks models and so on. To forecast the national macro economic using the system, and it brings a satisfied result in macro economic data forecasting. The relative error of the results is below 3 percent, and it lays a foundation of comprehensive integration system in the network environment.