针对社交网络中协同过滤推荐算法的推荐速度计算问题，提出了一种基于最近邻方法的改进计算方法，并对算法有效性进行了分析。该算法对用户的相似性度量采用基于最短路径的信任关系，用分层图和动态规划的方法进行计算，并在社交网络的应用中对关系链的深度进行限制。对该算法基于KDD Cup 2012 Track 1的数据进行了仿真，并与其他方法做了性能比较。实验表明，改进算法可以很好地平衡推荐效率和准确率指标。%In order to increase the speed of collaborative filtering recommendation in social networks, an improved nearest-neighbor algorithm is proposed in this paper. The proof of its correctness is also given in detail. The similarity measurement between users is based on trust relationship by using shortest path method. Layered graph and dynamic programming are applied to calculate the similarity. Furthermore, the recommendation speed can also be improved by limiting the depth of relationship chain in practical applications of social networks. The comparative simulations are carried out based on the KDD Cup 2012 Track 1 datasets. The results show that the better balance between the accuracy and the recommendation efficiency can be achieved by the proposed algorithm.
展开▼
