A Study On Predicting Item Difficulty In A Chinese Reading Comprehension Test Based On ANN

Item difficulty is one of the important item parameters. Usually, it is obtained by field-test, which faces a variety of problems in practice. Due to its own characteristics, Artificial Neural Network (ANN) is widely used in prediction research of various fields. This dissertation applies it to predict item difficulty of reading comprehension items in Chinese. This method does not need examinees: only through analyzing reading materials and stems can it predict the item difficulty of each item. Taking reading comprehension tests in HSK as training and test dataset, this article successfully predicts item difficulty in that significant correlations are obtained between the actual and predicted item difficulties.