Research And Application Of Trust Model Based On Social Network

Trust problems of commodities and trading partners gain more concern in case rapid development of e-commerce. We summarize the trust and reputation model researches; implement in-depth analysis of the application of credit model on the web. Then, a social network based, attack-resistant reputation and trust model is proposed in this paper. Reputation model is firstly established based on dealer scores, then trust is modeled using transitivity algorithm in a social network which is constructed by dealing relationships. Correction parameter is introduced to obtain subjective reputation value and objective trust value, which will help dealers make transaction decisions. The model uses two-dimensional cluster filtering to get rid of cheating data. The simulated experiments show that the model can properly describe the trust relationships between dealers, and performs well in resisting attack behaviors such as cheating feedback and shilling attack.There are difficulties to apply the model in commercial networks due to some problems such as division of labor, sparse network and clustering. An improved model is proposed in this paper with the introduction of similarity computation. This model discovers the relationships between clusters, and applies transmission of trust in the networks; finally predict trust value between two dealers. Through the experimental analysis with TAOBAO transaction data, we found our algorism has a very good improvement and complement to the existing reputation system of TAOBAO, and our model will simulate the interpersonal relationships of trust and recommendation in real life.