%0 Journal Article %T 融合社交网络的物质扩散推荐算法<br>Mass diffusion on coupled social networks %A 邓小方 %A 钟元生 %A 吕琳媛 %A 王明文 %A 熊乃学< %A br> %A DENG Xiao-fang %A ZHONG Yuan-sheng %A L(¨overU)Lin-yuan %A WANG Ming-wen %A XIONG Nai-xue %J 山东大学学报（理学版） %D 2017 %R 10.6040/j.issn.1671-9352.0.2016.030 %X 摘要： 在互联网信息推荐应用中,恰当地结合用户的社交信息能够进一步提升推荐的精度。 以用户为枢纽节点将社交网络和用户-商品二部图融合为耦合网络,并在此基础上提出了一种基于物质扩散动力过程的推荐算法,该算法将社交网络的朋友信息和用户选择商品的信息进行有机集成,是经典物质扩散算法的一种拓展。 在真实数据集Friendfeed和Epinions上的实验表明,在只计算小度用户的推荐准确率时,该方法比经典的物质扩散算法分别提高了38.48%和9.17%;当测试集所占比例为80%时,对于所有目标用户,算法较经典物质扩散算法的推荐准确率分别提高59.05%和21.62%。 因此,社交网络信息的加入可以显著提高对小度用户的推荐准确度。<br>Abstract: In the Internet information recommendation application, combining the users social information into recommend systems may further enhance the recommendation accuracy. We transform the social network and the user-commodity bipartite network into a coupled network by considering users as hub nodes and then propose a recommendation algorithm based on the process of mass diffusion dynamics, which integrates the information of friends in the social network and the information of the users selection of items in the user-item bipartite network. It can easily be seen that our approach is an extension of the classical mass diffusion algorithm. Experiments on the real datasets, Friendfeed and Epinions show that the recommendation accuracy of small degree users is improved by 38.48% and 9.17% respectively by comparing our proposed method with the classical mass diffusion algorithm. When the proportion of probe set is 80%, the improvement on recommendation accuracy is 59.05% and 21.62% than that of the classical material diffusion algorithm for all target users. Therefore, the addition of social network information can significantly improve the recommendation accuracy for small degree users %K 物质扩散 %K 社交网络 %K 推荐算法 %K < %K br> %K mass diffusion %K social networks %K recommender systems %U http://lxbwk.njournal.sdu.edu.cn/CN/10.6040/j.issn.1671-9352.0.2016.030