闭回路采样的网络结点特征学习方法

发布时间：2018-03-24 01:10

  本文选题：网络嵌入　切入点：闭回路采样　出处：《小型微型计算机系统》2017年09期

【摘要】：近年来,由于网络数据规模膨胀而导致传统的网络挖掘模型效率低下的现象,使得网络嵌入模型成为当前社会网络分析的热点.不同于以往模型的随机采样方式,本文考虑闭合回路机制对结点采样序列的影响,提出一种闭回路采样的网络嵌入模型,能够将大规模网络中结点的结构特征映射到连续的、低维度的向量空间.这样学习到的结点特征向量能够更好地反应网络的真实结构特性,并且可以很容易地应用到网络数据挖掘的分类、推荐和预测等任务.本文选取3个真实网络数据集进行多标签分类和聚类的实验,并与多个最新的基准方法对比,结果验证了该方法能够学习到更好的结点特征向量.
[Abstract]:In recent years, due to the expansion of scale and network data led to the traditional network mining model is inefficient, making the network embedded model has become a hot topic in the social network analysis. Random sampling method is different from the previous model, this paper considers the influence mechanism of closed loop sampling sequence of nodes, this paper proposes a model of close loop sampling network embedding that can be a large network structure characteristic of node is mapped to the continuous, low dimensional vector space. This node feature vector to learn to better reflect the network structure characteristics, classification and can be easily applied to network data mining, recommendation and prediction tasks. This paper selects 3 real network data set of multi label classification and clustering experiments, and compared with a new reference method, results show that this method can learn better node Eigenvector.

【作者单位】： 武汉大学计算机学院;汉口学院计算机科学与技术学院;
【基金】：国家自然科学基金项目(61272277)资助 中央高校基本科研业务费专项基金项目(274742)资助
【分类号】：O157.5
，

本文编号：1656025