基于LDA主题特征的微博转发预测

Predicting Retweeting Behavior Based on LDA Topic Features

微博转发是微博传播的重要途径,也是研究微博信息传播、舆情监控的最关键问题之一。研究用户转发行为对信息传播分析、舆情监控和热点提取有很大帮助。然而,当前对微博转发行为的研究大多是在宏观层面,为了解决微观层面预测用户转发行为问题,在分析影响用户转发的各类因素基础上,首先构建了微博特征和用户特征,然后通过将LDA抽取的微博隐含主题特征,与微博特征和用户特征相结合建立起基于主题特征的微博预测模型。实验结果验证了该模型在微博转发行为预测的有效性。

Retweeting is both the key way for information propagation in microblog and one of the most important issues in information dissemination and public opinion monitoring research. It is significant and helpful to study user retweeting behavior for analyzing informa- tion diffusion, monitoring public opinion and hot extraction. However, most of the researches are focused on the macroscopic stratum of the microblog social network which are difficult to apply on areas such as recommending ads or message and predicting the diffusion path. To predict whether a tweet will be retweeted by a user at the microcosmic level, this paper first builds the microblog features and the user fea- tures based on the analysis of the factors that affect retweeting behavior, and then establishes a predicting retweeting behavior model by combining the latent topic features extracted via using the semantic model of latent Dirichlct allocation with the microblog features and the user features. The experiment results have verified the feasibility of this model based on topic features.