Abstract: A new multi-label classification algorithm——Multi-Label Cluster Tree was presented. This algorithm combined the characteristics of transitional single-label cluster tree, and used the "clustering by label information algorithm" to divide the two space recursively until the clusters are sufficiently simple by the feature paces and the clustering information paces. Experimental results proved that the Multi-Label Cluster Tree algorithm outperforms the some existing popular classification algorithm, and the ability of classification is stronger than one of the sorting.