基于双层贝叶斯模型的商品分类算法研究

With the rapid development of e-commerce, product information is increasingly complex, the correct label affixed to commodities, which means classification, has become the key point to data mining in commodities. Firstly, according to the characteristics and requirements of commodity texts, this paper gives the methods of segmentation of Chinese texts, removing stop words, and representing the commodity texts using VSM. After that, a double-layer Bayesian model with Jelinek-Mercer smoothing is proposed in this paper. Finally, experiments in this paper give the tests and analysis of the proposed model. Results show that the double-layer model can improve the performance from the aspects of precision and recall comparing to single-layer Bayesian model. User could make a good choice between single-layer and double-layer model depending on the actual situation.