Application Of Four-parameter Logistic Model In Computer Adaptive Examination Syste

Educational evaluation,namely educational measurement and evaluation,is one of the three major fields of modern educational science research.In a narrow sense,it is a process in which educators measure and evaluate the ability,comprehensive quality and personality of the educated via certainly means,methods and tools.The traditional teaching examination is an educational evaluation based on the Classical Test Theory.Due to the existence of measurement errors,the true ability level of the participants is not equal to the measured value,and it is increasingly unable to meet the needs of educational evaluation.The computerized adaptive testing(CAT)system based on the Item Response Theory came into being.CAT generally refers to the adaptive evaluation by computer mediated,aiming to provide each participant with the optimal evaluation,so as to accurately evaluate the ability level of the participants.It can select suitable items according to the ability of the subjects,so as to estimate the ability of the subjects faster and more robustly.CAT is required to have the following conditions for robust estimation of subject ability:highly fit the underlying IRT model of subject response behavior,efficient selection strategy and robust ability parameter estimation.Therefore,to reduce the test time and improve the accuracy of ability estimation,this paper chooses four parameter logistic model(4PLM)as the underlying model,optimizes the selection strategy and 4PLM maximum likelihood estimation method(4PLM-MLE)of CAT,and proposes the optimal ability corresponding selection method and 4PLM robust estimation method(4PLM-Robust).Firstly,this paper introduces the research status of adaptive testing and item response theory,compares the advantages of item response theory over classical measurement theory,and systematically introduces four item response models and main CAT algorithms.Then,the CAT selection strategy with 4PLM as the underlying mathematical model is optimized,the optimal ability corresponding method is proposed,and the display expression of the optimal ability value of the subjects is derived.Simulation experiments and empirical analysis show that this selection strategy can effectively improve the efficiency of selection and reduce the system burden.Note that not every subject has the same guess on the same question,which leads to overestimation or underestimation of the subject’s ability.Therefore,4PLM-Robust estimation method is proposed to estimate the ability parameters.Ideal experiments and simulation experiments show that 4PLM-Robust can effectively estimate the participant’s ability robustly.This paper studies the CAT based on the four parameter logistics model,optimizes the selection strategy of maximum fisher information and the estimation method of ability parameter,which can correct the overestimation or underestimation of subject’s ability caused by guessing or sleeping phenomenon,and make the ability estimation of subjects more accurate.At the same time,it has the advantages of convenient test,free time and high efficiency of topic selection.