顶点带属性网络链接预测的参数选择方法
发布时间：2018-12-19 09:42
【摘要】：链接预测问题在社会学、人类学、信息科学以及计算机科学等各个领域都受到了广泛的关注.在基于相似度的链接预测的方法中,Katz指标是一种重要的顶点相似度指标.鉴于Katz指标中参数的可选择性,提出了一种基于参数选择的顶点带属性网络的链接预测算法.Katz相似度指标是基于路径相似性链接预测结果评价指标,Katz相似度指标中参数的取值会直接影响到Katz指标预测的结果.由于顶点带属性网络含属性和拓扑双重信息,算法思想是结合顶点属性信息进行参数选择,可以通过调节Katz相似度指标中参数的值,使Katz相似度尽可能和属性相似度靠近,将顶点属性相似度信息融入Katz相似度之中,以期达到属性信息和结构信息的有机融合.实验结果证明了该算法可以得到较高质量的预测结果.
[Abstract]:Link prediction has received wide attention in sociology, anthropology, information science and computer science. In the method of link prediction based on similarity, Katz index is an important index of vertex similarity. In view of the selectivity of the parameters in the Katz index, a link prediction algorithm based on the vertex and attribute network based on parameter selection is proposed. The Katz similarity index is an evaluation index based on the path similarity link prediction results. The value of parameters in Katz similarity index will directly affect the prediction result of Katz index. Because the vertex with attribute network contains attribute and topology information, the idea of the algorithm is to select the parameters by combining the vertex attribute information. The Katz similarity can be as close as possible to the attribute similarity by adjusting the value of the parameters in the Katz similarity index. The vertex attribute similarity information is integrated into Katz similarity in order to achieve the organic fusion of attribute information and structure information. The experimental results show that the algorithm can get high quality prediction results.
【作者单位】： 扬州大学信息工程学院;南京大学软件新技术国家重点实验室;
【基金】：国家自然科学基金项目(61379066,61379064,61472344,61402395)资助 江苏省自然科学基金项目(BK20130452,BK2012672,BK2012128,BK20140492)资助 江苏省教育厅自然科学基金项目(12KJB520019,13KJB520026)资助 江苏省六大人才高峰项目(2011-DZXX-032)资助
【分类号】：O157.5
本文编号：2386748
[Abstract]:Link prediction has received wide attention in sociology, anthropology, information science and computer science. In the method of link prediction based on similarity, Katz index is an important index of vertex similarity. In view of the selectivity of the parameters in the Katz index, a link prediction algorithm based on the vertex and attribute network based on parameter selection is proposed. The Katz similarity index is an evaluation index based on the path similarity link prediction results. The value of parameters in Katz similarity index will directly affect the prediction result of Katz index. Because the vertex with attribute network contains attribute and topology information, the idea of the algorithm is to select the parameters by combining the vertex attribute information. The Katz similarity can be as close as possible to the attribute similarity by adjusting the value of the parameters in the Katz similarity index. The vertex attribute similarity information is integrated into Katz similarity in order to achieve the organic fusion of attribute information and structure information. The experimental results show that the algorithm can get high quality prediction results.
【作者单位】： 扬州大学信息工程学院;南京大学软件新技术国家重点实验室;
【基金】：国家自然科学基金项目(61379066,61379064,61472344,61402395)资助 江苏省自然科学基金项目(BK20130452,BK2012672,BK2012128,BK20140492)资助 江苏省教育厅自然科学基金项目(12KJB520019,13KJB520026)资助 江苏省六大人才高峰项目(2011-DZXX-032)资助
【分类号】：O157.5
【相似文献】
相关期刊论文 前1条
1 肖熙;于宝海;;对《SCDD》优化法参数选择的探讨[J];上海交通大学学报;1982年01期
，本文编号：2386748
本文链接：https://www.wllwen.com/kejilunwen/yysx/2386748.html
