Link Prediction In Temporal Collaboration Networks Based On Game Theory

With the increasing tendency of the integration of the global economy,regional and international cooperation is becoming inseparable.Against this background,we must stress that cooperation is very important.With the advancement of modern science,as an effective organizational form,academic team,adapting the requirement of the development of the technology in the knowledge economy era,has become the important supporter of the discipline construction,playing an important role in in various fields.For the team collaboration with the character of complex system,which could be depicted and analyzed from system architecture and functional characteristics by complex network theory.In recent years,link prediction in complex networks has attracted widespread attention in academia.Link prediction is an important task in complex network analysis,which can be applied to many real-world practical scenarios such as recommender systems,information retrieval,and marketing analysis.The traditional link prediction methods based on node similarity,which have low time complexity but its prediction accuracy need to be improved.In addition,the forecasting accuracy of the methods based on the classification is higher than the methods based on node similarity,however which usually require additional information,and how to deal with the class imbalance problem is still a challenging problem.At present,most research on link prediction is mainly confined to the case of static network,while the real network systems evolve dynamically over time,so temporal network research will become an inevitable trend.Traditional link prediction methods can not been directly applied to temporal network,how to effectively integrate topology information and timestamp information is the key to solve such link prediction problems.To solve this question,we propose a novel semi-supervised learning framework,which integrates both survival analysis and game theory,and multi-agent autonomous computing framework is introduced.First,we carefully define a ?-adjacent network sequence,and make use of time stamp on each link to generate the ground-truth network evolution sequence.Next,to capture the law of network evolution,we employ the Cox proportional hazard model to study the relative hazard associated with each temporal link,so as to estimate the covariate’s coefficient associated with a set of neighborhood-based proximity features.Finally,to compress the searching space,we further propose a game theory based two-way selection mechanism to inference the future network topology.We finally propose a network evolution prediction algorithm based on autonomy-oriented computing,and demonstrate both the effectiveness and the efficiency of the proposed algorithm on real-world temporal networks.In the experimental stage,firstly,the empirical analysis of DBLP data sets is carried out,and some conclusions are summarized from the point of view of missing edges and reconnected edges,which have a certain guiding role in the experiment.Then,four temporal network datasets in Stanford Network Analysis Platform(SNAP)are selected and four ?-adjacent network sequences are generated.Next,we consider these four ?-adjacent network sequences as real network evolution sequences and apply the proposed method to four initial static networks.Finally,we compare the performance of the proposed method with five supervised learning-based link prediction methods and three probabilistic model-based link prediction methods,and the validity and efficiency of the proposed method are verified on real-time temporal network datasets.The experimental results show that the proposed method is superior to other algorithms in accuracy and execution time.