Research On The Prediction And Propagation Of Signed Networks

Signed network is a new research object of complex network.Different from traditional complex network,signed network adds sign information to the edge of the network.Sign information can well integrate a large number of emotional interaction information generated in the information society,so as to help us understand the social system more deeply.Therefore,we should first ensure the integrity of edge sign information collected before carrying out signed network related research.However,in the face of the massive data generated by the rapid development of information,the collection and sorting of information will inevitably lead to the loss of edge symbol information.How to use the existing network information to mine and predict the unknown edge sign information,so as to help us build a complete signed network is of great significance.Researchers have made abundant research achievements on Sign prediction algorithms.At present,classical sign prediction algorithms have achieved high accuracy,but the algorithm complexity is very high.With the rapid development of big data,the increase of complexity will cause huge time and space costs.In order to better balance the relationship between sign prediction accuracy and algorithm complexity,Paper proposes a sign prediction algorithm based on structural balance theory and status theory to calculate node similarity,and carries out experiments on multiple different data sets and test sets with different proportions.Compared with classical signed prediction algorithm based on common neighbor(CN)and link prediction in signed networks based on similarity and structural balance theory(PSNBS),the accuracy and complexity of the algorithm are analyzed.Our algorithm is very close to the classical algorithm in terms of prediction accuracy,but it is obviously superior to the classical algorithm in terms of algorithm complexity,and the time complexity is one order of magnitude lower than the classical algorithm.Based on the work of sign prediction,the dynamic propagation in signed network is also the key problem that people pay special attention to.In the classical propagation model,it is assumed that the contact probability and node threshold between nodes are homogeneous,while in the real society,the propagation between nodes will be affected by many factors.Therefore,in order to better describe the real propagation behavior,a new heterogeneous contact probability is proposed on the basis of node degree,and the definition of node threshold is proposed on the basis of structural balance theory.The improved contact probability and node threshold are introduced into the classic SIR propagation model,and the threshold propagation model based on signed network is established.In order to study the influence of edge sign on propagation,the improved model is simulated on symbolic networks with different negative edge ratios.It is found that with the increase of negative edge ratio,the average threshold value of nodes in the network increases continuously,the scale of propagation decreases continuously,and the propagation time increases significantly.In addition,this paper also studies the influence of fixed threshold and heterogeneous threshold on propagation,and finds that the fixed threshold propagation model has little change in propagation scale and speed,while the heterogeneous node threshold has a more significant impact on propagation,which can more effectively reflect the propagation behavior in the real world.