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.
Proper representations are the key to the success of most machine learning and data mining algorithms.Due to the expansion of the network scale and the low efficiency of traditional network mining model,many network embedding models have been proposed in recent years.Different with the random walk sampling method of the previous models,this paper proposed a novel network embedding model based on loop sampling,which maps each vertex of a network to a fixed-length,low-dimensional vector.Vertices of the graph with similar semantics are assigned with similar vectors.Such low-dimensional vector representations naturally encodes the structure of a network can be easily fed into machine learning models in a various range of network analysis tasks,such as classification,link prediction and visualization.Experimental results on three real-world datasets show that our model outperforms the other state-of the-art baselines on multi-label classification and clustering tasks.
出处
《小型微型计算机系统》
CSCD
北大核心
2017年第9期1940-1944,共5页
Journal of Chinese Computer Systems
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
