Abstract: By calculating the document frequency of each feature in a category and other categories,an existing feature selection algorithm-DPM (Discriminating Power Measure) compares the contribution of features for one category with other categories,and extracts the features that reveal larger differences among categories.On the basis of the research of DPM,a new feature selection algorithm is proposed,which considers the importance of word frequency of each feature in the calculation of discriminating power of features.Experiments demonstrate that the proposed algorithm has a better performance.