Music style classification becomes more and more application in commercial audiencerecommendation. But any simple and efficient method has not been found for solving the great deal of music online. In this paper we developed a framework based on Naive Bayesian Network (NBN) to perform music style classification. We acquired such important features after deeply studying, and construct our specific NBN for music style classification. The results of the experiments in our method indicated that the simplicity and accuracy of music style classification improved greatly.