Neural network-based cognitive diagnostic method

Abstract

Statistics cognitive diagnostic methods were complex and could not judge cognitive bugs very well. To solve this problem, a hybrid method combining principal component analysis, self-organizing feather map and probabilistic neural networks was promoted. It was applied in cognitive diagnostic. The data was got from 488 students of high school who took in Chinese language test. The results showed the principal component analysis could reduce the dimensions for SOM input data, and get the cognitive attributes. SOM network could divide the subjects into categories, and identify the cognitive shortages of different categories. Probabilistic neural network could judge cognitive bugs of the new students accurately. It is a valuable cognitive diagnostic method.