Research on drug side effects contributes to reducing health risks for patients and decreasing drug development costs. In recent years, machine learning methods have emerged as prominent tools to support analyzing and predicting drug side effects. In this paper, we propose a clustering model based on knowledge graph embedding (GEBC) to predict potential side effects. GEBC integrates various drug features into a heterogeneous network, then mines the underlying relationship of nodes in network using Node2vec algorithm. Finally, GEBC clusters the node embeddings by Gaussian Mixture Model, and builds a Bayesian probability prediction model. GEBC achieves 0.4343, 0.442, 0.411 AUPR score on three benchmark datasets, respectively.