预测高考考生能力水平 调控高考试题难度研究探新

Exploration of Predicting the Ability of College Entrance Examinee and Adjusting the Difficulty of College Entrance Examination

在我国,社会各界普遍希望高考历年分数线基本保持不变,但是由于每年题目不同,考生能力水平也有波动,要实现上述目标就需要预测考生能力水平并据此调控试题难度。本研究基于课题组2010年起在海南和云南试点的高考等值研究,根据项目反应理论对外锚卷进行题目参数的估计,结合条件最大似然估计和同时校准的方法,使各试卷的能力水平均置于同一量尺。再以回归和K近邻法建立外锚卷和高考能力水平间的预测关系,并以2014年的高考数据进行检验。结果显示,在英语和数学(文、理)上,回归方法对2014年考生能力水平的预测精度都较高,K近邻法仍需进一步改进。

The stakeholders of College Entrance Examination （CEE） feel that admission scores should remain unchanged. Predicting the examinees＂ ability and adjusting the difficulties to keep the admission score stable is a great challenge for testing institutions because of that every examination is brand new and the mean ability level of examinees varies every year. Based on the teams＂ equating research about CEE in Hainan and Yunnan provinces since 2010, the study includes the following aspects： a）estimating the parameters of items and persons on external test and CEEs with concurrent calibration and conditional maximum likelihood to transform the scores from different tests onto a common scale, b）the prediction relationship was built between external test and CEE with regression model and K Nearest Neighbors （KNN） according to the past data and verified using the data of CEE in 2014. The results show that the regression model had higher prediction accuracy than KNN on English and Math subjects.