融合多类型信息的社交网络用户表示学习方法

A Method of Multi-Type Information Embedding for Social Network User

网络表示学习方法进行社交网络用户表示可以避免大量的特征工程,同时方便对不同类型的特征进行融合。由于社交网络通常规模大且数据类型丰富,采用基于神经网络的网络表示学习方法,融合社交网络中的多类型信息学习用户表示,提出了一种融合多类型信息的社交网络用户表示学习方法。对社交网络用户涉及的文本、网络结构和属性标签信息设计了独立的神经网络结构和目标函数,并通过对目标函数求加权和的方式进行融合,采用梯度下降算法进行联合优化得到社交网络用户表示。Cora和Weibo数据集上的实验结果表明:所提方法可以更好地融合社交网络中的多类型信息,获得更有区分度的用户特征,可有效提升分类任务的准确率。

User representations in social networks can avoid complicated feature engineering and facilitate the integration of different types of features. Because social networks are usually large-scale and rich in data types,a method of multi-type information embedding for social network users is proposed, which is based on neural network representation learning method to learn user representations by integrating multiple types of information in social networks. Independent neural networks and objective functions are designed for text,network structure and attribute information of social network users. The models are fused by the weighted sum of the objective functions. The gradient descent algorithm is used for joint optimization. The experiments on Sina Weibo data set and public data set show that the proposed method can better integrate multi-type information in social networks,obtain discriminative user features and effectively improve the accuracy of classification tasks.