基于D-S证据理论的网络表示融合方法

随着网络表示学习技术的发展，在网络结构信息的基础上，越来越多的研究者考虑融入额外辅助信息来提升网络表示效果.针对现有网络表示学习方法中对于多属性特征融合缺乏冲突判别与评价指标的问题，本文在已有研究基础上，提出一种基于D-S证据理论的网络表示融合方法.本方法首先通过SVM算法给出不同属性信息对融合表示结果的支持度，然后利用证据组合规则计算网络表示学习中的融合评价指标，并依据混淆矩阵考虑各类别在节点分类中的局部可信度.在3类数据集上的仿真实验表明：本方法对于检测网络表示融合中的冲突，提高表示融合效果具有一定的指导意义.

Method of Network Representation Fusion Based on D-S Evidence Theory

With the development of network representation learning technology,researchers are increasingly considering the integration of multi-dimensional attribute information to enhance the performance of network representation.In view of the lack of conflict discrimination and evaluation index for multi-attribute feature fusion in existing network representation learning methods,this paper proposes a network representation learning fusion method based on D-S evidence theory.Firstly,the support degree of different attribute information to the fusion result is given by SVM algorithm.Then,the fusion evaluation index in network representation learning is calculated by using evidence combination rules,and the confidence degree of each attribute is fully considered based on confusion matrix.Simulation results on three types of data sets show that the method can effectively detect conflicts in network representation fusion and improve the performance of fusion representation.