The link prediction method that only considers topology information without considering node labels does not achieve a good prediction result for label networks. The paper proposes a link prediction method combining Node Labels with Common Neighbors (NL-CN) to solve this problem. First, a similarity index based on node labels (NL) is defined. The similarity of two nodes is measured by the cosine of the angle between the label feature vectors of the two nodes. Secondly, the NL index and the common neighbor index are combined to obtain the binding index and link prediction for social networks using the binding index. Finally, experiments are conducted in six label networks, and the experimental results show that the method can effectively improve prediction accuracy. In the Gene, Citeseer and Cora networks, the AUC values of the NL-CN index were improved by 10.84, 4.76, and 0.22 percentage points, respectively, compared to the traditional Cos+ index with better performance.