With the rapid development of Internet and the improvement of the storage techniqu es, more and more music resources are got ten , and correspondingly their classification becomes an important issue . Up to now, three main classification methods have been proposed , which are Tag - based approach, content - based approach and the methods based o n machine learning. Although t hese approach es have been analyzed and compared based on the music features; there are still no deep research es on the intrinsic relationship between the music . This paper focus es on this issue and presents a novel m usic c lass ification a lgorithm based on both the music c ontent and the critical e igenvector c entrality . Firstly , the new algorithm extracts the MFCC ( Mel Frequency Cepstrum Coefficient ) and Rhythm features of music, and then the K - Means algorithm is used to cluster t he se feature s. A s a result, K - cluster centers are generated . C orrespondingly , the music network using similarity relations can be constructed . At last, the music network is analyze d by using a graph - based e igenvector c entrality and the music classification is completed by joint considering the K - cluster center s. Moreover, experiments are provided to evaluat e the new proposed classification algorithm , and its performance is also compar ed with the conventional content - based classification algorithm . The resul ts show that the new music classification algorithm is more reasonable to satisfy th e needs of most of users.