社交网络中的链路预测方法改进  

Improvement of Link Prediction Method in Social Networks

作　　者：李旗旗 徐敏[1] 

机构地区：[1]南京航空航天大学计算机科学与技术学院,江苏南京211106

出　　处：《计算机技术与发展》2017年第11期37-40,共4页Computer Technology and Development

基　　金：国家"973"重点基础研究发展计划项目(2014CB744900)

摘　　要：社交网络在近些年得到了迅速发展,如今各个行业都在努力加入社交元素,如何提高链路预测方法在社交网络中的预测准确度成为一个热门研究方向。链路预测方法由于网络结构的不同会表现出不同的预测效果,因此可以根据社交网络的结构特性对链路预测方法进行改进,从而提高在社交网络中的预测准确度。社交网络是对人与人之间某种社会关系的描述,因此和其他复杂网络相比,会表现出独特的网络性质和结构,其中最主要的是"小世界"特性和无标度特性。针对社交网络的这种特性,对原有的链路预测方法进行改进,在共同邻居方法的基础上加入了优先连接对节点相似性的贡献。真实社交网络数据集的对比实验结果表明,改进后的方法在没有增加时间复杂度的情况下提高了预测准确度。The social network has been developing rapidly in recent years. Various industries are now trying to integrate social elements, so how to improve the accuracy of link prediction methods in social networks has become a popular research. Due to the different network structures, the link prediction methods will be different in prediction performance so that it can be improved according to the characteris- tics of social network structure,improving of the accuracy of prediction. The social network is a description of certain social relations between people, so compared with other complex networks,it will exhibit its unique properties and network structure, of which the most important is the ＂small world＂ and scale-free characteristics. According to the characteristics of social network, the previous link prediction methods can be improved, adding the contribution of priority connection based on common neighbors. The experiments on real social net- work data sets show that the improved method can improve the accuracy of prediction without increasing time complexity.

关 键 词：社交网络 链路预测 网络结构 无标度网络 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]