In the past few years, the volume of junk information on the internet has grown tremendously, researchers' begun to handle this issue. In this paper, DCM (Discriminative Category Matching) algorithm is employed to filter the web information according to the content of information. To our knowledge, the algorithm is the first introduced into filtering. It takes the relative importance of a feature in a category, across categories and the average importance of a feature in a category into account, and modifies the traditional TF-IDF term weighting scheme to compute the domain class central vector. Experiments indicate that the algorithm is suitable for information filtering.