Application Of Data Mining On Producing Process Of Stream Model Industry

Data mining, also named as KDD (Knowledge Discovery in Database), is an advanced process, in which we can pick up many trustful, novel, useful and readable patterns from very large amounts of data. Classification is one of the most important branches of data mining research works. Target of classification is to find out a classification function or model. The model can map a single record in database to a pre-assumed class. Classification can be used to forecast. The output of classification is discrete class label. Based on different input vector, forecast relevant output vector.In Stream Model of Industry, The application of Data Mining must bring higher product yield and quality, and create more benefit. Fuel Alcohol Ltd. Company of JiLin Province(shortened form: JFA) has WINCC(by SIEMENS developed for data collected of product line), Office Automatization system and Import-Sale-Storage system running. During investigating on the spot, discussing with technology engineers on every station, based on the customer request report, and in order to efficiently use function and data structures of existing system, we determined to use J2EE platform and Data Mining technology developed Business Intelligence System for JFA --- typical stream model of industry. Fuel Alcohol Ltd. Company is the biggest one of development and providing business, which is the corn biochemistry industry and the burgeoning energy sources, it is the first large-scale producing fuel alcohol base that is authorized to build by Chinese government.Business Intelligence system is integrated with Office Automation (OA). In order to realize work without papers, and resolve the inefficient work flow problem that papers fly all over the sky, reports are a heap of. When developing BI system, uses the technology of Data Mining on real time data analysis that is correlative with quality control of product line and import-sale-storage data. Combining with real circumstances, considering multifarious reports, entire system is make up of four function module, in order to automatically make diversified daily report, mensal report, quarter report, annual report, and use these reports to statistic analysis and forecast.Firstly, numerous real time producing data is irrelevant, usability is very poor, so it must be associated, be formed data structure in favor of analysis and mining. BI system uses the AIP-- Application Integration Platform distilled production data into own data warehouse, and clean the data noises for the next analysis and forecast;Secondly, use universe of BI system begins to define the relation among data in data warehouse, name dimension and measurement with business glossary that customer is used to apply. Which makes customers avoid learning the inside structure of database and complicated database glossary; Customer uses the defined universe and applies report tools, based on different function of every department, automatically makes needful daily statistic analysis reports of all departments, and can drill at different layer to finish the analysis and description of history data;Thirdly, BI system uses the history data in data warehouse and applies Neural Network----- Sense Machine evaluating arithmetic to works out the relation between product quality and parameter obtained scope of every phase on product line, during learning arithmetic of two hundred instances to get steady measure parameters, in order to make strong reliability and high credibility classified pattern; Lastly, BI system can forecast the parameter of every phase by feasible classified pattern, and effectively control product quality.BI system has formally been devoted to application in JiLin Fuel Alcohol Ltd. Company, its running status is good, and satisfy customer's function request. About the complicated evaluating arithmetic of evaluated pattern, I will continue to research with the increase of customers' request, in order to resolve the most practical problems.