Abstract In this paper, the feature selection of subjective and objective classification of Chinese micro blog has been studied. For the features in Chinese micro blog, the combination of lexicon and statistics is used to extract candidate features. By this method, eight candidate features are extracted. And a feature selection algorithm based on rough sets and probability-weighted is proposed. Using the algorithm, six features are selected. The experiment result shows that these features selected by the algorithm achieve good result in subjective and objective classification of Chinese micro blog.