With the rapid development of e-commerce and community forums,the number of product reviews has increased dramatically. Review data has become an important data source for enterprises to improve the quality of goods and services. Reviews include user’s attention to product features and corresponding emotional tendencies.Therefore,in order to mine the emotion distribution of users on multiple features of commodities in a fine-grained way,a comment classification model based on comment feature words and a feature emotion classifier based on Naive Bayes algorithm were proposed in this paper. From the user’s review,this paper analyzed the emotional orientation of the product’s multiple features.