Wireless networks provide a realm in which cooperation among large numbers of egoists can be attained. As the non-cooperative behaviours of nodes will significantly degrade the network performance, effective cooperation incentive of nodes has become a hot issue in cooperative communication. Although a reputation system can stimulate nodes to cooperate with each other, the recommendation-based trust model may cause a decline in network performance, because of fake recommendation, convergence of iteration and node redemption. To address the above issues, a topology transform-based recommendation trust model is proposed to relieve the malicious effects on the accuracy of recommendation trust, which stem from fake recommendation. Some mathematical analysis and simulation results reveal that the global trust of this proposed trust model has convergence characteristic, and the reputation system with this model can more effectively stimulate node cooperation and performs better in terms of packet successful delivery ratio and mean number of packets dropped. Moreover, the topology transform-based recommendation trust algorithm makes this reputation system more suitable for large-scale networks.