基于网络节点文本增强的链路预测算法

已有的链路预测算法主要是基于目标网络结构信息的,没有考虑到与目标网络相关的文本信息。针对此问题,提出一种基于网络节点文本增强的链路预测算法。将网络节点的文本内容融入到网络表示学习过程中,使学习得到的网络表示向量中含有节点的文本属性。通过余弦相似性算法构建出目标网络的相似度矩阵。在3个真实的数据集上做链路预测仿真实验。实验结果显示,相比于现存的多种链路预测算法,该算法预测结果的精确度有明显提升,同时能够有效且准确地挖掘网络中节点间的结构关联性和内部相关性。

LINK PREDICTION ALGORITHM BASED ON TEXT ENHANCED OF NETWORK NODES

Existing link predictions algorithms are mainly based on the target network structure information and do not take into account the text information related to the target network. In order to solve this problem, we proposed a link prediction algorithm based on text enhanced. The text content of network nodes was integrated into the learning process of network representation. It makes the learning network representation vector contain the text attributes of nodes. Then, the similarity matrix of the target network was constructed by the cosine similarity algorithm. The link prediction simulation experiments were performed on three real data sets. The experimental results show that compared with the existing link prediction algorithms, the accuracy is improved significantly. The algorithm also can effectively and accurately mine the structural correlation and internal correlation between nodes in the network.