Abstract: Aiming at the problem of trust between buyers and sellers in online shopping, the paper establishes the subjective reputation model based on recommendation. The model not only introduces objective factors such as transaction time, transaction amount and trust recommend credibility, but also introduces subjective factors such as the first impression. The simulation experiments based on Matlab show that the [Zi→j] model can guarantee transaction success rate of more than 0.8 especially in large-scale deal, moreover it can better against malicious nodes（especially intermittent malicious nodes）of malicious behavior than other trust models, and it has  high stability.

Key words: online shopping, direct trust, recommend trust, trust recommend credibility, malicious node

摘要： 针对网络购物中买卖双方的信任问题，建立了基于信誉推荐的主观信誉模型。该模型不仅引入了交易时间、交易金额及信誉推荐可信度等客观因素，还引入了第一印象等主观因素。基于Matlab的仿真实验表明，[Zi→j]模型能保证交易成功率的成功率在0.8以上，而且[Zi→j]模型比其他的信誉模型能更好地对抗恶意节点（特别是间歇性恶意节点）的恶意行为，具有较高的稳定性。

关键词: 网络购物, 直接信誉, 推荐信誉, 信誉推荐可信度, 恶意节点