The accurate identification of tumour subtypes in the treatment of tumors is important; the classification of different tumor subtypes has recently received a great deal of attention in the field of bioinformatics. The paper sorts genes using Relief algorithm and selects the initial subset of the genes of tumor information firstly. Then, calculates the weighted gene sets using the forward attribute reduction algorithm based on neighborhood rough set model. Then the weighted K-NN algorithm is used to analyze the data in order to detect differentially expressed genes. The results showed the feasibility and effectiveness of the method proposed in this paper.