结合用户偏好和相似性的网络结构推荐算法

Network structure recommendation algorithm based on user preferences and similarity

网络结构推荐算法在资源分配过程中,仅判断用户是否选择过项目,未考虑用户显式偏好对资源分配的影响,导致推荐项目单一,为此提出一种结合用户偏好和相似性的网络结构推荐算法。在二部图网络结构推荐的基础上,引入用户显式评分,采用最大最小值方法标准化用户不同评分作为网络结构的权值,在资源分配过程中综合考虑项目的度以及用户间相似性对资源分配的影响。实验结果表明,该算法与其它算法相比明显提高了平均排序值(ranking score)和海明距离,提高了推荐的准确性和多样性。

For the single issue of recommended items using network structure recommendation algorithm,because the algorithm only determines whether the user selects over projects or not,ignores impacts of user preferences on resource allocation,a network structure recommendation algorithm combining user preferences and similarity was proposed.On the basis of the bipartite network structure algorithm,introducing user＇s preference scores,max-min method was used to standardize user＇s different ratings as the weights of network architecture.In the resource allocation process,the degree of project and impacts of user similarity on the resource allocation were considered.Experimental results indicate that comparing with the other algorithms,the improved method obviously increases the average ranking value and Hamming distance,improving the accuracy and diversity of recommendation.