It’s difficult to extract the development trend of tourist and viewpoint for tourist attractions, propose the tourist attractions viewpoint analysis method based on machine learning multi-level constraints CRF model. Analyzes common review analysis method, use CRF algorithm to extract review from data set, and establish an automatic recognition model for false reviewers and their viewpoints based on the characteristics of the reviewer relationship, filter false reviewer viewpoints, and combine viewpoints with respect to the relevance of the review viewpoints. The experimental results show that, the method is better than other common methods in terms of precision and recall, and the PR and ROC curves are good, the method has a good decision-making function for the tourist to choose the traveling target.