高考命题中试题难度预测方法探索

Research on Predictive Methods of Item Difficulty in the College Entrance of Examination

运用命题教师主观评估、多元线性回归分析和BP神经网络建模三种预测方法,对高考命题过程中试题的难度进行预测,并对三种方法的预测性能进行比较。结果发现,三种预测方法均具有较高的预测准确度,其中,BP神经网络预测模型对试题难度的预测准确度相对更高,误差相对更小。

A predictive method of item difficulty was constructed via teacher＇s subjective evaluation, regression analysis and neural network modeling respectively. The optimum method stands out after com- parison of the three kinds of predictive methods. The conclusions indicated that the three methods all have accuracy to prediction of item difficulty. According to the indications of relative errors, absolute errors and average errors, the BP neural network showed a better result of the prediction of item difficulty.