Abstract: This paper proposed an improved global harmony search algorithm (IGHS) to solve the shortest path (SP) problem. Firstly, a dynamical genetic mutation probability was defined and introduced into the harmony search algorithm, which can effectively prevent the IGHS from trapping into the local optimum. Secondly, a dynamical priority-based encoding approach was used for harmony representation in IGHS, and a path will be built according to the value of decision variable in the harmony vector. The shortest path will be obtained through updating harmonic memory. The experiments have been carried out on different network topologies for networks consisting of 20~100 nodes, and three performance indexes were used to compare the proposed algorithm with PSO-based and traditional harmony search (HS) based shortest path searching algorithm. It is shown that the performance of the proposed algorithm surpasses PSO-based and HS-based approaches for this problem.