Commodity Classification in Hierarchies

In e-commerce transactions, goods are classified according to the hierarchical structure, which refers to a tree category. In the process of classification, we shall consider the special features. While using brand name for category, for instance, the degree of distinction characteristic of brand is higher. Based on this, we prepare a dictionary of brands for Chinese words segamentatin on one hand and use a kind of "discriminative naive Bayes classifier" model on the other hand. According to our experiment, we can get a result that the Naive bayes classification model is better than standard bayesian one.