Abstract:

The recommendation algorithm based on social information effectively alleviates the data sparsity and cold start problems in the recommendation algorithm, which has attracted great attention in recent years. However, social information still has the problem of data sparsity, and the binary data provided by social networks can not measure the degree of trust between different users. Therefore, first of all, it uses random walk with restart to obtain important nodes in social network. Then, the important nodes trust propagation algorithm is proposed to establish the trust relationship between indirect user nodes and important nodes. At the same time, the trust weight between users is further quantified by using the structural information of nodes to get more accurate recommendation results. Experiments on three open datasets show that the social recommendation combined with important nodes trust propagation enriches social information and effectively improves the accuracy and recall rate of the recommendation algorithm.

Key words: recommendation algorithm, social information, important nodes, trust propagation, Random Walk with Restart（RWR）

摘要：

融合社交信息的推荐算法有效缓解了推荐算法中的数据稀疏性问题和冷启动问题，近年来受到极大的关注。但社交信息依然存在数据稀疏性问题，而且社交网络提供的二值数据无法衡量不同用户间的信任程度。针对这些问题，利用重启随机游走算法获取社交网络中的重要节点。提出重要节点信任传播算法建立重要节点和其他用户节点之间的信任关系，同时利用节点的结构信息进一步量化用户间的信任权重，以得到更精确的推荐结果。在三个公开数据集上的实验表明，结合重要节点信任传播的社会化推荐算法（INTP-Rec）丰富了社交信息，有效地提高了推荐算法的准确率和召回率。

关键词: 推荐算法, 社交信息, 重要节点, 信任传播, 重启随机游走算法