Prediction Of Microblog Heat Based On Multi-dimensional Feature Fusion

After released in 2009,microblog has evolved into a widely accepted self-media,now it's the most timely and convenient channels for people to get information.Besides,microblogging is the most centralized network platform where people distribute information,discuss social hot spots and express their personal point of view.Microblog is the network platform on which people most intensively post information,discuss social hotspot,and express their personal point of view.Studies on microblog users' forwarding behavior are of great significance in public opinion monitoring,business marketing,behavior prediction,etc.User's forwarding behavior is influenced by a wide range of facts,including their own behavior characteristics,feature of content,and the network environment where there are placed in.We proposed a method to forward prediction.1.Microblog forwarding feature extraction.People has different emotions for a same microblog,Some people may behave positively,others may behave negatively.We based on the emotional characteristics of microblog from the perspective of individual people to measure the influence of microblog bloggers.And we extract inherent features such as the number of followers,retweets,and comments to predict the popularity.2.Microblog topic feature extraction.According to the influence of microblog content characteristics on dissemination,and considering the influence of the subject's bias,the LDA document theme model was used to model microblog content to form a text topic description.We use the topic vectors and other relevant features extracted by the LDA model as the main features of the microblog content level to analyze the popularity3.Microblog social network graph construction.For the microblog blogger's own information,we obtained the attention relationships between all relevant users to construct a relationship network,which is represented by a graph.4.Hot microblog forwarding prediction.Based on the characteristics of microblog and the social network,we analyze the two aspects of retweet breadth and retweet depth,based on neural network and other algorithms to constructing a dual-branch structure to predict the hot microblog.Our experiments show that our method performs well and effectively improved the microblog forwarding prediction results.