To recognize the DNA sequence and mine the hidden information to achieve the classification of organisms are viewed as a difficult work to biologists. As we known, the amino acids are the basic elements to construct DNA. Hence, if the codon usage of amino acids can be analyzed well, the useful information about classification of organisms may be obtained. However, if we choose too many amino acids to perform the clustering analysis, the high dimensions also lead the clustering analysis to be a complicated structure. Hence, in this study, we will incorporate the dimension reduction and clustering technique into proposing a novel approach, and it will provide the useful information about organisms classification based on the codon usage. Finally, we also apply a case including eighteen bacteria to demonstrate the rationality and feasibility of our proposed approach.