基于三分支神经网络的多特征微博传播预测模型

MULTI-FEATURE PREDICTION MODEL OF WEIBO PROPAGATION BASED ON TRIPLET NEURAL NETWORK

针对现如今微博传播预测的模型考虑因素不够全面的问题,提出基于三分支神经网络的多特征微博传播预测模型。该模型以三分支神经网络结构为框架,利用LDA(Latent Dirichlet Allocation)模型提取微博文本特征,利用改进后的PageRank算法分析用户影响力特征,并与微博是否带有图片、链接和视频等其他特征相融合。经实验验证,该模型在微博传播预测准确度上较已有双分支模型有显著提高,且稳定性良好。

Nowadays,there are many models for the prediction of Weibo propagation,but the factors are not completely comprehensive.To solve this problem,this paper proposes a multi-feature prediction model of Weibo propagation based on triplet neural network.The basic framework of this model was a triplet neural network structure.In this model,LDA model was used to extract the text features of micro-blog,and improved PageRank algorithm was used to analyze the characteristics of user influence.The model combined with other features such as whether the microblog had pictures,links and videos.Experimental results show that the proposed model significantly improves the accuracy of Weibo propagation prediction compared with two-branch models,which has good stability.