This paper presents a novel method to predict popularity of artists on music streaming services based on network analysis. In the proposed method, a network whose nodes correspond to artists can be constructed by collaboratively using audio features of their audio tracks, textual features of their biographies and social metadata representing related artists. In addition, the proposed method constructs a classifier that can predict whether each artist is popular or not by extracting features considering structure of the obtained network. Experimental results on multiple real-world datasets constructed by using one of the largest music streaming services show the effectiveness of the proposed method.