Analysis And Prediction Of Weak Ties In Social Networks

As the scale of social networks continues to expand,the information contained in the network has been explosively increasing.How to efficiently extract valuable information from the network has drawn many researchers' attention.Link prediction predicts new links that may be formed in the future based on the current and historical structure of the network.It can mine hidden information from the network and promote the evolution of the network structure.At present,link prediction has a wide range of applications in commodity recommendation,social security,biomedicine etc.While the accuracy of link prediction has gradually increasing with the continuous development of research on link prediction,the utility of the inferred new links is rarely concerned especially when it comes to information diffusion.Inspired by weak tie theory,this thesis defines a special type of links named bridge links based on the community structure(overlapping or not)of the network.In sociology,bridge links are usually regarded as weak ties which play a crucial role in information diffusion.This thesis also defines the utility of links in information diffusion based on average shortest path.Considering that the accuracy of most link prediction methods is high in predicting strong ties but not much high in predicting weak ties,this thesis proposes a new bridge link prediction method named iBridge,which aims to infer new bridge links using biased structural metrics in a PU(positive and unlabeled)learning framework.The experimental results in 3 real online social networks show that iBridge outperforms several comparative link prediction methods(based on supervised learning or PU learning)in inferring the bridge links and meantime,the overall performance of inferring common links is not compromised,thus verifying its robustness in inferring all new links.In addition,this thesis defines another type of weak ties named NECLink based on network representation learning.Since network representation learning can learn the low-dimensional space vector representations of the nodes,the spatial distance between nodes can be calculated to measure the strength of the link.This thesis proposes to cluster the nodes according to their embeddings and then the links between the clusters are defined as weak ties.This thesis verifies the role of NECLinks in information diffusion on three real social network datasets.The experimental results show that NECLink have higher utility than common links and even higher utility than bridge links in most cases.Experiment also compares the utility of NECLinks in different network representation learning method settings,which verifies the stability of NECLink extraction method.