基于贝叶斯网络的数据挖掘及其在新药试生产中的应用

《基于贝叶斯网络的数据挖掘及其在新药试生产中的应用》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、上海交通大学硕士学位论文基于贝叶斯网络的数据挖掘及其在新药试生产中的应用姓名：朱娟娟申请学位级别：硕士专业：计算机应用技术指导教师：张申生20060101上海交通大学硕士学位论文摘要注的未知条件的结果为用户提供决策支持最后在论文结尾以具体的实验数据展示了该套方案的可行性以及有效性关键词数据挖掘贝叶斯网络期望最大值算法联结树算法系统聚类法第II页上海交通大学硕士学位论文ABSTRACTTHEDATAMININGBASEDONBAYESIANNETWORKSANDAPPLYINGITTONEWMEDICINES’PREPRODUCTIONABSTRACTDataMiningisaprocessw

2、hichpickingoutusefulbutpotentialinformationandknowledgefromhugeandstochasticdata.BayesiannetworkscomefromBayesianstatistics.Theyaregraphicalmodelsthatcangraphicallypresentthejointprobabilitydistributionfunctionsbetweenarraysofvariables.Recently,theresearchersandscholarsfoundthatapplyingBayesiannetw

3、orksintodataminingcandigmultilayerandmultipointcausalrelationships.ItisagowithoutsayingthingthatcombiningthedataminingwiththeBayesiannetworks.Thispaperintroducesthemainfunctionsofthenewmedicines'preproductionmanagementsystembriefly,alongwithstudyingthewholeprocessofdataminingbasedonBayesiannetworks

4、emphatically.Aimingatthecharacteristicsofthedata,wemakesomeimprovementstotheexistingBayesiannetworks'learningandinferringalgorithms.Atlast,weapplytheimprovedalgorithmsintooursystemandalreadygetsomegoodresults.Thedataofthenewmedicines'preproductionisdiscontinuous,incompleteandnothuge.Thesecharacteri

5、sticsbringsomedifficultiesforBayesiannetworks'structurelearning.Firstly,weneeddefineandlookupsimilardatainordertofindouta第III页上海交通大学硕士学位论文ABSTRACTcertainextenthugedatafromthedatabase.Secondly,wedealwiththesedatausingsystemclusteringmethod,whichisoneoftheclusteringanalysemethodsandcandispersethedisc

6、ontinuousdata.Thirdly,accordingtotheclassicalEMalgorithmofBayesiannetworks'learning,weintroduceanewStructEMalgorithm,whichcansolvetheproblemofincompletedataandcanbesomewhatsatisfiesthereal-timeneeds.Fourthly,aimmingattheBayesiannetwork,wecanenactmentsomeknownconditionsandthenuseJunctionTreealgorith

7、mforinferring.Throughinferringwearchivethepurposethatisgivingouttheunknowninformationandprovidingdecision-makingsupportfortheusers.Attheendofthepaper,Ipresentthefeasibilityandvalidityofthewholesystemthrough