An automatic web page classification is needed for web information extraction, but the number of keywords of web pages is so giant that many classifications are not speedy or capable of self-learning. In this paper, a fuzzy classification method for web pages, which is based on fuzzy learning and parallel feature selection, is proposed. Fuzzy learning of parameter c{ik} is adopted to increase the accuracy, while parallel feature selection based on weighted similarity is used not only to decrease the dimension of the features but also to let parameter σ{ik} need no learning. The weights of features are deducted in theory, and to speed up the calculation of weights, a parallel sum algorithm of the matrix is proposed.