基于链路预测的潜在科研合作关系发现研究
发布时间：2019-05-05 17:30
【摘要】：[目的/意义]从科研合作网络"小世界现象"和"无标度特性"出发,对网络中的未知链路产生连接的可能性进行预测。[方法/过程]采用Katz指标和cosine距离计算网络中的路径相似性和研究者科研兴趣相似性;在此基础上,构建加权预测算法,并采用AUC指标和F1度量确定节点相似性在算法中所占的比重。[结果/结论]通过实证研究发现,在路径相似性基础上,引入一定权重的节点相似性,可以达到较好的预测效果。[局限]在节点相似性计算时,仅考虑了研究者科研兴趣因素,暂未考虑研究者地理位置对合作关系的影响。
[Abstract]:[aim] from the point of view of "small-world phenomenon" and "scale-free characteristic" of scientific research cooperation network, the possibility of connection of unknown links in the network is predicted. [methods] Katz index and cosine distance were used to calculate the similarity of paths and researchers' interest in scientific research. On the basis of this, the weighted prediction algorithm is constructed, and the proportion of node similarity in the algorithm is determined by using AUC index and F1 metric. [results / conclusion] through the empirical study, it is found that on the basis of path similarity, a certain weight of node similarity can be introduced to achieve a better prediction effect. [limitation] when computing the node similarity, only the researcher's interest in scientific research is considered, and the influence of the researcher's geographical position on the cooperative relationship is not taken into account for the time being.
【作者单位】： 江苏大学科技信息研究所;
【分类号】：G353.1;O157.5
，
本文编号：2469774
[Abstract]:[aim] from the point of view of "small-world phenomenon" and "scale-free characteristic" of scientific research cooperation network, the possibility of connection of unknown links in the network is predicted. [methods] Katz index and cosine distance were used to calculate the similarity of paths and researchers' interest in scientific research. On the basis of this, the weighted prediction algorithm is constructed, and the proportion of node similarity in the algorithm is determined by using AUC index and F1 metric. [results / conclusion] through the empirical study, it is found that on the basis of path similarity, a certain weight of node similarity can be introduced to achieve a better prediction effect. [limitation] when computing the node similarity, only the researcher's interest in scientific research is considered, and the influence of the researcher's geographical position on the cooperative relationship is not taken into account for the time being.
【作者单位】： 江苏大学科技信息研究所;
【分类号】：G353.1;O157.5
，
本文编号：2469774
本文链接：https://www.wllwen.com/kejilunwen/yysx/2469774.html
