In most graph-based semi-supervised methods, graph structure is often set in advance, which leads to the fact that the algorithm can’t learn an optimal graph in the process of label propagation. Therefore, this paper proposes a method called Adaptive Graph-based Semi-supervised Learning Method (AGSSLM). This method can learn the optimal graph and label simultaneously by using the iterative optimization method. Moreover, this method can also obtain higher classification accuracy with fewer labeled samples. The experimental results validate the effectiveness of this method.