Abstract: Cancer recognition with gene expression profile was studied. Due to large redundant and noise information in the gene expression data and the sensitiveness to selected feature genes,the classification is lack of generalization capability. With studying the gene expression profiles, two cancer recognition models including global component model (GCM) and cancer component model (CCM) were constructed. And a weighted voting strategy was applied to propose an Co-classification Approach based on GCM and CCM for cancer recognition (CAGC). Test experiments were conducted on Leukemia, Breast, Prostate, DLI3CL, Colon and Ovarian cancer dataset respectively, and great performance is acquired by CAGC on all datasets. The experiment results show that the recognition solution and the generalisation are strengthened by combination of GCM and CCM.

Key words: Gene expression profile, Cancer recognition, Global component model, Cancer component model