Abstract-Multi-label classification is an important area of data mining, in where decision tree is one of the effective means to solve the problem. It faced a huge challenge of performance caused by large size of data. First, we translate the multi-label classification to several binary classifications. Then we analyzed the potential parallelism of decision tree based multi-label classification algorithm from four parts and overall applied them in the training and predicting phases. The parallel algorithm was implemented with MPI and the performance of parallel decision tree based multi-label classification algorithm is analyzed and compared program designations and experiments, which demonstrate that our parallel algorithm could improve the computing efficiency and still has some extensibilities.