Aiming at the problem of Influence Maximization (IM) in heterogeneous information networks, an Influence Maximization algorithm (DAGIM) based on Directed Acyclic Graph (DAG) was proposed. Firstly, the influence of nodes was measured based on the DAG structure, and then the marginal gain strategy was used to select the nodes with the most influence. The DAG structure has strong expressive power, which not only describes the explicit relationship between different types of nodes, but also depicts the implicit relationship between nodes, and more completely retains the heterogeneous information of the network. Experimental results on three real datasets verify that the performance of the proposed DAGIM algorithm is better than those of Degree, PageRank, Local Directed Acyclic Graph (LDAG) and Meta-Path-based Information Entropy (MPIE) algorithms.