基于Word2vec和粒子群的链路预测算法

Link Prediction Algorithm Based on Word2vec and Particle Swarm

链路预测中普遍存在两大问题:特征提取困难和类别数据不平衡.本文借鉴文本处理中的深度学习特征提取算法和优化问题中的粒子群算法,提出一种基于词向量的粒子群优化算法(Word2vec-PSO).该方法首先通过随机游走产生网络序列后,利用Word2vec算法对节点序列特征提取.然后在有监督的条件下,利用粒子群算法对提取好的特征进行筛选,并确定重采样的参数来解决类别数据不平衡问题,并分析了不同链路预测算法的计算复杂性.最后将本文的算法与基于相似性、基于深度学习、基于不平衡数据的3类链路预测算法,在4个不同的时序网络中进行实证对比研究.结果表明,本文提出的链路预测算法预测精度较高,算法更加稳定且具有普适性.

There are two major problems in the link prediction:The difficulty of feature extraction and the imbalance of class data.In this paper,an algorithm based on word vector is proposed by using the deep learning feature extraction algorithm in text processing and the particle swarm optimization algorithm in the optimization problem.The method firstly generates a set of node sequences through random walks,and uses the Word2vec algorithm to extract node sequence features.Then,under the supervised conditions,the particle swarm algorithm was used to filter the extracted features,and the resampling parameters were determined to solve the imbalance problem of category data.It also analyzes the computational complexity of different link prediction algorithms.Finally,the algorithm of this paper is compared with three link prediction algorithms based on similarity,deep learning,and unbalanced data,and empirically studied in four different time series networks.The results show that the link prediction algorithm proposed in this paper has more accurate prediction accuracy and is more stable and universal.