基于NSGA2的网络环境下多标签种子节点选择

发布时间：2018-01-17 15:28

  本文关键词：基于NSGA2的网络环境下多标签种子节点选择　出处：《电子与信息学报》2017年09期 　论文类型：期刊论文

  更多相关文章： 社交网络 多标签分类 NSGA 系统开销

【摘要】：随着社交网络规模的不断扩大,网络节点的标签分类也不再单一,变得丰富多样,这些促使了社交网络中的多标签分类问题成为一个重要的研究领域。以前的研究重点主要集中在提高预测网络节点标签的精度上,而忽略了得到节点信息所产生的包含时间消耗和计算资源等在内的系统开销问题。可现如今随着网络规模不断扩大且复杂性不断增强,之前所忽略的系统开销问题变得越来越严重,增加了预测标签的成本,加重了预测网络节点标签的难度。该文针对这一问题提出了基于NSGA2算法的网络环境下多标签种子节点选择算法(NAMESEA算法),目的是在能大大降低预测节点标签所消耗的系统开销的前提下一定程度上提高预测标签的精度。该文将NAMESEA算法与其他多标签预测算法在多个真实数据集上进行实验对比,结果证明NAMESEA算法大大降低了预测节点标签的系统开销并且提高了预测精度。
[Abstract]:With the continuous expansion of the scale of social networks, the label classification of network nodes is no longer single, becoming rich and diverse. This makes the classification of multiple tags in social networks become an important research field. The previous research focuses on improving the accuracy of the prediction network node labels. However, the problem of system overhead caused by getting node information, including time consumption and computing resources, has been neglected. But now, with the increasing of network size and complexity, it is becoming more and more complex. The problem of overhead that was previously ignored has become more and more serious, increasing the cost of forecasting tags. In order to solve this problem, a multi-label seed node selection algorithm based on NSGA2 algorithm is proposed in this paper. The purpose of this paper is to improve the accuracy of prediction tags on the premise of greatly reducing the system overhead consumed by prediction node tags. In this paper, the NAMESEA algorithm and other multi-label prediction algorithms are used in multiple real numbers. An experimental comparison was made on the set. The results show that the NAMESEA algorithm greatly reduces the system overhead and improves the prediction accuracy.
【作者单位】： 合肥工业大学计算机与信息学院;科学技术部基础研究管理中心;路易斯安那州立大学计算机与信息学院;
【基金】：国家973规划项目(2013CB329604) 国家重点研发计划项目(2016YFB1000901) 国家自然科学基金项目(61503114)~~
【分类号】：TP18;TP393.09
【正文快照】： (2)(科学技术部基础研究管理中心北京100862)(3)(路易斯安那州立大学计算机与信息学院拉斐特70503美国)近年来随着社交网络应用的发展普及,社交网络吸引了越来越多学者的研究目光[1-5],其中一个重要的研究方向就是社交网络中的多标签预测问题[1]。利用标签预测我们可以通过网，

本文编号：1436843