Multicasting is a communication service that allows an application to efficiently transmit copies of data packets to a set of destination nodes. The problem of finding a minimum cost multicast tree can be formulated as a minimum Steiner tree problem in networks, which is NP-completeness. MPH (minimum path cost heuristic) algorithm is a famous solution to this problem. In this paper,we present a novel solution TPMPH (two phase minimum path cost heuristic) to improve the MPH by generating the nodes and the edges of multicast tree separately. The cost of multicast tree generated by the proposed algorithm with the same time as MPH is no more than that of MPH in the worst case. Extensive simulation results show that TPMPH can effectively improve the performance on MPH, and performs better in large-scale networks and wireless networks.