Link Prediction Based On Node Contribution And Path Influence

With the rapid development of the digital and information age,people are immersed in the complex system of interconnected everything.Among them,various associations of complex systems can be simulated and characterized through the nodes and edges of complex networks.Link prediction,as one of the research hotspots in complex networks,can predict the possibility of unknown connections through known information,and has gradually been widely applied in many fields.In social networks,the likelihood of strangers becoming friends can be predicted based on existing friendship relationships,thereby making friend recommendations.The link prediction algorithm based on network topology has received widespread attention due to its easy to obtain network structure and strong interpretability.However,this type of algorithm considers more the topological properties of a certain aspect of the network,and the similarity differentiation between node pairs is low,which limits the prediction accuracy.In response to the problem of insufficient information extraction in the link prediction process,this article analyzes the topological structure information of complex networks from two aspects: node contribution and path influence,and studies similarity algorithms that fuse node resource transmission and path information.The main work of the thesis is as follows:(1)Propose a link prediction algorithm based on node resource transmission differentiation and global path influence.Based on the different contributions of neighboring nodes to end nodes,assign larger weights to common neighboring nodes with small degree values,and design a node resource transmission capacity quantification module.Subsequently,all reachable path information between nodes is extracted through the global path module,and finally,the two modules are linearly fused for link prediction.The experimental results on 10 real datasets show that the fusion of resource transmission capability and global path can improve the accuracy of link prediction.(2)Design a link prediction algorithm based on node resource transmission differentiation and local path influence.In order to balance the prediction accuracy and operational efficiency of the algorithm,a local path influence module is used to replace the global path influence module.Firstly,the first-order neighbor loop problem in the third-order path acquisition process is improved.Then,a local path influence module is proposed based on node clustering coefficients.Finally,the local path influence module is integrated with the resource transmission differentiation module,and adjustable parameters are used to control the effect of local path influence.The experimental results show that adding a local path influence module can effectively reduce the time complexity of the algorithm.(3)Extract multi-dimensional network structure information and study the impact of local path influence of clustering coefficients on the common neighbor algorithm.Extract multi-dimensional information from different algorithms,fuse the common neighbor module and clustering coefficient local path influence,and set adjustable parameters to adjust the impact of clustering coefficient local path influence weight on the algorithm’s prediction results.The experimental results show that adding clustering coefficients to local path influence can improve the prediction performance of the algorithm.