3)No Creation and self-learning of DSS. It is only of the There are three levels of CP network which are importfunctions which programmed by developer. arrangement of ideas, argue unexpectedly arrangement of ideas and output arrangement of ideas. The self-configuration 4)Bad ability of real time. reflecting neural network is composed of import arrangement In order to solve the above problems of DSS, the bionics of and argue unexpectedly arrangement. The basic argumentneural network and the intelligence of mining system are made network is composed of argue arrangement and outputuse of to explore the solutions in theory. arrangement. The argue arrangement of ideas which is between the import arrangement of ideas and output arrangement of III. FUNCTION OF NEURAL NETWORK TO DSS ideas reflects the statistics characteristics of import mode and output mode. Then there is a reflecting between import mode The function of neural network to DSS is as following by and output mode through argue arrangement of ideas. CPmeans of analyzing of characteristic of neural network and network has been widely used in lots of fields such as modesproblems of DSS. The learning function, the parallel clustering, statistics analysis, data compressing and so on.distributing processing function with large scale, non-lineardynamics with constant time and the collectivity of neuralnetwork are made use of to have a realization of automation ofknowledge learning, self-learning of natural languageprocessing system, overcoming the difficulties of “assembledblast” and “infinite recursion”, adaptive parallel associatingreasoning, promotion of deciding ability of DSS and processingof real time. As shown in Fig. 2, a intelligent decision support system ofneural network is composed of knowledge, data and model. Itis of main four sub-systems, neural network, reasoning system,data mining system of neural network and natural languagealteration system.  Figure 3. Configuration Drawing of CP Network There are two kinds of data mining which are directly data mining and indirectly data mining. The aim of directly data mining is that a model is built up by means of available data and a special variable is described. The definition of indirectly data mining is that no some real variable is described with a model but some kind of relationship has been built up.  B. Reasoning of Neural Network Figure 2. Block Drawing of IDSS of Neural NetworkA. Data Mining of Neural Network Data mining of neural network is a data mining mode basedon neural network technology. There are five basic tasks ofdata mining, relative analysis, clustering, concept description,error monitoring and forecast. The forward feeding neuralnetwork, for example BP network, is usually worked in concept Figure 4. Sketch Map of Repeat Reasoning of Neural Networkdescription and forecast. The counter spreading (CP) neuralnetwork can be worked in statistics analysis and clustering. As The main research of neural network system is the doubleshown in Fig. 3, it is a configuration drawing of CP network directions reasoning method based on the data driving and aimwhich created by American neural calculating expert Robert driving of neural network. Reasoning is the main method ofHecht－Nielsen solutions. The course of knowledge reasoning is the process of solutions. There are some problems such as “assembled blast”
and “infinite recursion” in the traditional reasoning method. A. Man-machine Alteration SystemThe parallel processing of neural network is the best method of The alteration structure has been established by man-solving above problems. An explanation of reasoning course of machine alteration system which has input and output betweendouble directions associate memory (BAM) network is shown system and user. Man-machine alteration system is thein the next. As shown in Fig. 4, the first arrangement of repeat important part of IDSS which is of all functions as shown inreasoning of neural network is of no calculate function, but of Fig. 5.fan-out function which is distributed the output to input. Aninput vector A is applied up to power matrix and an outputvector B is turn out. And vector B is applied to the turn matrixＷＴ of power matrix W, then a new output vector A is turnout. The course is repeated until a steady point of networkwhich A and B are constant. The steady point is called ashomeostasis.  There is a formula of the repeat course as following. B=F(AW) (1) A=F(BWT) (2) A is output vector of the first arrangement, herein, B isoutput vector of the second arrangement, W is the powermatrix between the first arrangement and the second, and F ispower function.  Equation (1) can be used to fulfill the reasoning of datadriving and Equation (2) can be used to fulfill the reasoning of Figure 5. Configuration of IDSS Based on Neural Network and Data Miningaim driving. Double directions reasoning can be realized byhomeostasis of BAM. The homeostasis is the crossing point ofdata driving and aim driving of BAM. And it is the decision B. Neural Network, Data Mining, Solutionssolution. Neural network, data mining and solutions include two modules which are solutions module and data mining module.C. Natural Language Alteration System of Neural Network Data mining module works up in order to gain knowledge The main research of natural language process (LS) is the needed through making use of the model of models base,syntax analysis and meaning analysis based on neural network. method of methods base and knowledge of knowledge base.Natural language is belonged to non-numerical valve symbol Solutions module works up in order to configure or half-which is symbol flow with different numbers. It is of its own configure the problems through making use of thesyntax and means system and its data structure, means corresponding model of models base, method of methods base,expression and calculation rules are rather different from knowledge of knowledge base and data of data base. Reasoningnumerical valve information. The core of natural language can be made use of for the non-configuration problems.processing system of neural network is how to understand theknowledge and the expression of natural language. The basic V. APPLICATION OF IDSS BASED ON NEURAL NETWORKtasks of syntax analysis system based on neural network are (1) AND DATA MINING IN USING OF ENERGY AND PROTECTION OFconfirmation of syntax structure of input sentence, which is a RESOURCESidentification course based on neural network, (2)standardization of syntax structure, which is a conclusion There is a lack of natural gas and oil in China. And there iscourse that lots of input structures turn into a few of input a great air pollution of coal burning. There is a kind of newstructures according to some syntax exchanging relationships. energy，so far, biological energy , that is grain alcohol, for the substitute of oil and natural gas. But grain alcohol is made from grain. Therefore, it is not suitable to develop grain alcohol in IV. IDSS SUPPORTED BY NEURAL NETWORK AND DATA stead of oil and natural gas. Some experts suggest that marsh MINING gas should be developed in order to replace the oil and natural IDSS supported by neural network and data mining is gas.shown in Fig. 5. It is derived from the combination oftraditional DSS with data mining technology in order to For the above problems, we make a research on whetherincrease the intelligence of system. It is composed of man- marsh gas can be made use of in stead of oil and natural gas bymachine alteration system based on neural network, data means of IDSS based on neural network and data mining.。mining, reasoning and solutions, data base management, The main researching movements are as following.knowledge base management, methods base management and 1) Man-machine Alteration System which is ofmodels base management. convenience to users.
2) Building up data base which is composed of oil, natural neural network technology into IDSS. And there are lots ofgas, coal, grain alcohol petrol, cost of marsh gas, using valve, problems which should be studied deeply in the future.pollution, cost of over pollution, health cost and so on. 3) Building up knowledge base which derived from the REFERENCESexperts.  HOU Fu-jun, WU Qi-zon, “Prediction of Length Sequence of Railways 4) Building up reasoning system based on neural network in Operation Based on Genetic Algorithm and Simulated Annealingand data mining which can fulfill the repeat reasoning function Algorithm Optimized Neural Networks”, Journal of Beijing Institute ofand turn out the homeostasis point of system and make junior Technology, 2004,(03)decision.  TIAN Fu-qing, FENG Chang-lin, LIU Jun, “Evaluation of integrated supportability for naval gun weapon system based on BP neural 5) Building data-revised system which can fulfill the repeat networks”, Journal of Naval University of Engineering, 2007,(03)reasoning function again by self-motion when a new data is  FENG CHANGLIN TIAN FUQING, “C3I System Efficiencyadded and turn out a new homeostasis point of system and Evaluation based on BP Neural Networks”, Microcomputer Information,make the last decision. 2007,(29)  YANG Jie, YAN Qing-dong, “Algorithmic Choice and Betterment in the BP Network of Cane Sugar Crystallization”, Microcomputer VI. CONCLUSION Information,2008,(12) It is creation that data mining and neural network has been  LIU Pei-feng, ZHANG Wen-bin, WANG Qi, “Study on the Locationused in the study of IDSS. There is a great foreground of the and Segmentation Algorithm of Vehicle License Plate with Gray Image”, Microcomputer Information, 2008,(18)realization of IDSS by means of the combination of knowledge  ZENG Ming, WEI Yan, “Research of Intelligent DSS based on Neuralpattern and neural network pattern and the introduction of Network”, Microcomputer Information, 2008. (24)