基于偏好相似度的混合信任推荐模型

Preference Similarity-Based Mixed Trust Recommendation Model

针对P2P网络中可信数据不完整的问题,提出了将局部可信度与全局可信度相结合的基于偏好相似度推荐的混合信任模型(Preference Similarity Recommendation Trust,PSRTrust),借助相似随机游走策略修复稀疏的可信度矩阵;对不合理假设呈现power-law分布进行合理化改进;并给出了可信数据的分布式存储和计算的分布式方法。仿真实验表明,PSRTrust模型有效地提高了在可信数据不完整情况下的交易成功率,并且在遏制恶意节点影响上有一定提高。

A large number of new nodes in peer-to-peer network make the trust matrix sparseand data insufficient. This leads to inaccurate global trusts of peers which are computed by trustmatrix iteration and finally the low success rate of transaction. PSR-Trust, a mixed trust modelcombining global trusts and local trusts based on preference similarities, restores sparse trust ma-trix by Similarity Random Walk. This model optimizes the unreasonable assumption, gives a dis-tributed implementation method and data storage. Mathematic analyses and simulations resultsshow that the proposed model is more robust under general conditions where malicious peers co-operate in an attempt to deliberately subvert the system, and the success rate of transactions ishigher compared with current trust model.