a new graph-search algorithm based on multi-attribute decision making (MADM) is proposed. Firstly, A* algorithm is used for graph-search, but when cost functions of two nodes are equal, three attributes describing a node are introduced. Secondly, a multi-attribute decision based on maximum deviation principle is used for nodes evaluation in order to select the best node for expanding. The proposed algorithm could overcome deviation brought by node evaluation based on information theory metrics only, which results in high accuracy. The outcome of simulation verification at the end of this paper manifests that this algorithm has excellent accuracy as the exhaustive algorithm, and is more quickly for large scale computation.