The flooding of biological data such as genome and protein data derived into insilico approach has become a prediction tool useful for the scientific community. At present, the function of such proteins is determined by the experiments that can be time consuming and costly. So there is a need of a tool for determining the function of the proteins. The prediction parameter of the protein that can be recognize the protein /non-protein classification, features such as amino, dipeptide composition and PSSM Profiles. In this study, support vector machine (SVM) algorithm is used to classify oxygen binding proteins with six main classes through amino acid, dipeptide, and position specific scoring matrix (PSSM) profile. We achieved the maximum accuracy was 83.45% (MCC 0.77), 82.69% (MCC 0.76) and 89.20% (0.85) of AC, DC and PSSM respectively. All models were developed using non-redundant dataset having (50%) cutoff of 1526, 1545 positive and negative sequences. The five fold cross validation has been applied to evaluate the performance. Our experimental result shows that our approaches are faster and achieve generally a better prediction performance over the existing method.