Neural networks based attribute hierarchy method in cognitive diagnosis

The neural networks based attribute hierarchy method (NN-AHM) is promoted in this study, it could be applied in cognitive diagnosis. AHM could classify examinees´ responses into a set of attribute mastery patterns. Very large sample is needed for AHM, and it is hard to describe the examinee master degree for each attribution. The generalized regression neural network is used in this study to estimate the attribute probabilities in small sample to provide information about the examinee´s attribute-mastery level.