Research On Node Prediction In Bipartite Network

At present,research on node prediction in networks mainly falls into two categories: one is the positioning of the "source node" in the network;the other is the discovery of "hidden node" in the network.However,there is a lack of research on the prediction of new nodes.In response to the current situation,by collecting the papers in the journal and the corresponding keywords information,the bipartite network is weighted and projected into a keyword relational network,and the generation of new paper nodes is predicted by using the keyword combination on the network.The main research contents are as follows:(1)The interaction between keyword nodes can be divided into two types,one is similarity,which is used to indicate the possibility of keywords appearing together.This thesis assumes that the information is propagated in a random walk mode on the keyword weighted network.The two-way information propagation quantity takes the geometric mean to express the similar average intensity between nodes,and constructs the local random walk index using geometric mean(LRWGM).Then this index is used to predict the new paper node prediction experiment,and the results show that this method is more advantageous than the traditional indicators;(2)The other is the proposed node mutual exclusion(ME),which can be used to describe the difficulty level of the common occurrence of keywords.For example,two keywords with the highest connotation are rarely present at the same time.This thesis uses the deviation degree between the co-occurrence times and the average occurrence times of two keywords to measure the mutual exclusion between nodes,which is helpful to describe the interaction relationship between nodes more comprehensively;(3)In the case of considering the similarity and mutual exclusion between nodes at the same time,this thesis believes that there is a closer relationship between nodes with similarity greater than mutual exclusion,and uses this condition as an important basis for selecting keyword combinations,which is based on the establishment of bipartite networks.The node prediction algorithm with mutual exclusion(NPME)is compared with the node prediction algorithms based on the similarity index,and the results show that this method can predict the occurrence of new papers more accurately.