A Topological Social Information Dissemination Model

The problem of information dissemination in social networks is one of the important topics of social networks in recent years.The traditional social information dissemination models are based on known conditions for the user's connection in the network to predict the trend of information dissemination,but in real social networks,generally,the structure of social networks is not known.Even in the real world,the topology of social networks is often not available.In this case,the process of simulating information dissemination is a challenging task.This paper proposes two non-topological social models,namely social information dissemination model based on user influence(NT-II)and interest and GRU-based social information dissemination scope prediction model(NT-GI).The first model uses a representational learning approach to construct two hidden spaces.One space is called user-influenced space.Each user is mapped to a vector in the space.The distance between vectors represents the influence between users.Another space is called the user interest space.Each user and each propagation item is mapped into a vector in the space.The user's preference for the propagation item is inferred from the distance between the corresponding vectors.The closer the distance is,the more the degree of affection is.Big.In this paper,a gradient descent method is used to give an effective learning algorithm for the model parameters.Experimental results on real data show that the model can more accurately simulate the propagation process and predict the propagation results,and it is superior to the existing non-topological structure propagation model in multiple metrics.The second model is a topology-free information propagation range prediction model.The algorithm uses GPU neural network training vectors.Because there is no topological structure and there is no input user propagation sequence and corresponding user vectors,we propose two kinds of selection.The method of propagating sequences,and two methods for vectorizing user nodes,these vectors will be used as the input vectors of the neural network model.After the learning of hidden layers,the vector parameter values are automatically obtained.Finally,through the attention mechanism,each group of vectors is integrated and weighted,and then the MLP mechanism can predict the scope of social network information dissemination.Compared with the prediction accuracy of other model propagation range,this model has better prediction effect.