基于IRT模型的題庫智能組卷策略.pdf

1、江西師范大學碩士學位論文基于IRT模型的題庫智能組卷策略姓名：李佳申請學位級別：碩士專業(yè)：計算機應用技術指導教師：丁樹良20070401AbstractOneoftheimportantapplications011ComputerAssistedInstruction(CAI)isComputerBasedTesting(CBT)，assemblingtestantomatieallynotonlyisthebaseofCBT，but

2、alsoistheimportantcomponentofCAIThemostimportantkeyonthetestconstructioncou僦ishowtosolvetheseverelyconstrainedproblemsrsuchascontentbalancing，typebalancing，total伊叫eete)Ancwtestconstructionstrategymaximizingtheconstraints

3、weightedinformation(MaxCWDstrategybasedonItemResponseTheory(IRDisproposedinthispaperComparedwiththeLinearProgramming(LP)testeonstrucfionsWategyMaxCWIstrategyhassuchadvantages，(1)Itcanbypassthecomputationalintensityandalg

4、orithmiealsophisticationofthemathematicalprogrammingapproaches，thecomplexityoftimecomesdownfrom研)toO(n2l092玎)，SOitcallbeeasilyimplementedinthecurrenttesteonstruetionautomatically(2)SimulationsshowthatMax—CWIstrategyisver

5、yeffectiveinseverelyconstraintmanagementunder13Severeconstraints，iteBnproductmanytestswhichconformtothetestblueprintatthes鋤etimebycomputerthesetestsarecalledsuccessfulteststheitempool’Sutilizationratioisover63％Thisviewpo

6、inthash’tbeenmentionedininternalperiodicalsatpresent(3)AnewconceptiontheaveragetestinformationbywhichtheestiIn啦accuracyofeverytraitpomofexamineeisproposedinthisthesisTheresultsofMonteCarlostudyshowthatthecorrelationcoeff

7、icientbetweentheaveragetestinformationandthetestinformationisveryhighAtthesametime，anfwproblemarisesthesesuccessfultests’qualitiesaredifferentanditisunfairtotheexamineesfromdifferentbatchesonthesametypetestanditaffectsth

8、eassessmfntoftbeexaminee’SreliabledegreeseriouslyToconquertheilewproblemseveralotherMaxCWImethods，suchasa_straitifiedtestconstructionstrategyb—slraltifiedtestconstructions仃ategyandrandomMax_CWlstrategyareproposedinthispa

9、per皿eyalsoCanproductcopiesofsuccessfultestsandCalleffectivelyaveragethetotaltestinformationamongtheall—successfultestsSotheyCanimprovethetests’qualities，especiallythebehavioroftheiandonlMax_CWIstrategyisbestKeywords：Max_