Abstract: As the previous evaluation methods of evolving models can not be credible because of its inconformity, in this paper, we propose a new method by applying the theory of link prediction and maximum likelihood estimation. Based on the Internet autonomous system networks, we find that GLP is better than Tang which is not agreed with previous results. Moreover, the optimal parameters of these two evolving models are different from the original ones. The experimental results support our optimal parameters with which the corresponding models can generate more real networks.