Feature selection is one of the key steps in text classification. To some extent, it can affect the performance of text classification. In this paper, we firstly proposed an optimized document frequency-based word frequency and document frequency and then presented the feature resolution based on the optimized document frequency. Meanwhile, we introduced rough set into feature selection and provided an attribute reduction algorithm based on the correlation matrix of equivalence classes. We finally put forward a feature selection method combining the presented feature resolution with the provided attribute reduction algorithm. The proposed feature selection method firstly employs the presented feature resolution to select some valuable text features and filter out useless terms to reduce the sparsely of text feature spaces, and then it uses the provided attribute reduction algorithm to eliminate redundant features. The comparative experimental results show that the proposed feature selection method has certain advantages in consumed time, macro-average, micro-average, and average classification accuracy.