Document Classification Based on Neural Network

This paper investigates a new approach to the use of the citation links to refine the initial results obtained from the content-based classifier. The goal was successfully achieved by developing a neural network based method to refine the class labels of the documents obtained by the content-based naive Bayes classification method. The new method was examined and compared with the content based Naive Bayes method on a standard document classification data set with increasing training set sizes. The results suggest that the new approach can significantly improve the system performance and that the citation link information is important for scientific document classification. While the neural network method has achieved good results.