The Research And Application Of Generalized Partial Credit Model Of CAT

In recent years,with the development of the computer network technology, it gradually appeared a lot of adaptive systems based on item response theory,they overcomed the old setting test paper mode and test method of classical test theory.The more mature adaptive system is for dichotomously scored items, which means there are only two kinds of answer,right or wrong,it is far from satisfying the requirements of polytomously scored item,just for the feedback is limited.Therefore,the adaptive system for polytomous items has important practical value and realistic significance.In the thesis, the background,current situation and the theoretical model of the polytomously scored item adaptive system are discussed based on three educational measurement theories.In combination with the characteristics of programming problem in C programming course,the thesis selects the Generalized Partial Credit Model (GPCM) whose step parameter is non monotone increasing as the theoretical model of adaptive system,and discusses in detail the GPCM,its item information function,ability estimation method and selection strategy.On this basis.the paper proposed a solution based on GPCM adaptive test system.According to the GPCM and the characteristic that the polytomously scored item has multiple step parameters,the paper constructs a database applying to GPCM and in-depth discusses discrimination related to the test,exposure and ability estimation method which is used in the test process.In order to reduce the calculated amount and improve the operating efficiency of system,before using the variable weighted item selection strategy,it gains the remaining item bank by using SQL which according to the related information that the student have had; to control the exposure of the system, it needs ascending sort the remaining item bank by discrimination and then uses the variable weighted itemselection to choos item from the sorted item bank. After the item was finished,it uses the MMLE/EM parameter estimation algorithm which has a easier convergence and the Newton-Raphson method to estimate the item parameters,the results provide the basis for realizing the dynamic adjustment of the parameters.This will make the item parameter more reasonable. At last, the author uses the J2EE technology to realize the programming questions adaptive test system based on the GPCM in Eclipse development environment.After the test experiments,by analyzing the outcome data,it proves that the variable weighted item selection strategy can better improve the selection efficiency and save time,and reflects that the system has the better accuracy and safety.lt can be better test the level of student’s acquired knowledge and real ability.