Information Diffusion Prediction Through Representation Learning

Information diffusion is defined as the process that information reaches new individuals through the interaction between users.The information diffusion prediction mainly predicts the users who participate in the information diffusion process.With the advent of the information age,information diffusion is becoming more and more common and complex.The prediction of information diffusion has also aroused people's attention.In this paper,a variety of representation learning methods are used to predict the information diffusion.The main work of this paper includes:1.The information diffusion prediction model based on knowledge representation learning is designed and implemented.We consider that information diffusion prediction can be transformed into link prediction problem in knowledge graph,and the related knowledge representation learning models are introduced to solve information diffusion prediction problem.In the proposed model,the likelihood probability is used to predict the information diffusion,and the limitations of the poor interpretability of the previous models in the prediction process are overcome.By introducing the concept of infected time weight,we quantify the influence of users on prediction results in information diffusion sequence,and solve the problem that evaluation part in previous models does not make full use of infected time of users.The experimental results show that our model proposed in this paper has a certain promotion effect on the effect of information diffusion prediction.2.Representation learning methods are applied to the initialization of relation vector and the quantification of user impact.Related language representation learning methods are used to initialize relation vector of our model.Network representation learning models and document similarity measure are used in the quantification of user influence.To sum up,with the help of representation learning,we avoid the tedious Feature Engineering,enrich and improve the physical meaning of the research object,and combine the related external features,so as to enhance the prediction performance of the information diffusion prediction model.