Minimum Steiner tree problem is a NP complete problem, and widely used in communication network point to multi-point routing. In order to realize more link sharing, reduce the cost of the desired Steiner tree, an algorithm named NWMPH (Node Weight based Minimum cost Path Heuristic) was proposed to solve the Steiner tree based on weighted node. The algorithm constructed a weighted formula of nonregular points, for each nonregular point weighting value. According to the weights of modifying the link cost. By modifying the cost shortest path in order to connect all regular points, get the minimum tree containing all regular points. For part of the data to calculate STEINLIB standard data set, the results show that: NWMPH algorithm and MPH algorithm used basically the same time. The cost of NWMPH algorithm to get Steiner tree is less than that of MPH algorithm. NWMPH algorithm uses less time and costs less to get Steiner tree than KBMPH algorithm.