With the development of e-commerce, blog, social networking sites and micro-blog become much more flourishing. The Internet has entered a new era, and the sentiment classification of online comments related to individual decisions, business management and also social security. A model of sentiment classification based on interval-valued intuitionistic fuzzy sets and the interval-valued intuitionistic fuzzy operator was proposed to calculate the interval-valued intuitionistic fuzzy numbers of feature words. Meanwhile, by using membership, non-membership and hesitation to quantitatively describe the feature words. It can get the sentiment tendency through the sentiment synthesis, in order to take more accurate analysis results of sentiment tredency. Finally, throughing the comparative experiment based on same corpus, it proves that this model has high feasibility, correctness and classification performance.