Research Of Spreading Mechanism And Social Recommendation Algorithm In Complex Networks

Research on biological and social spreading mechanisms on complex networks focuses on macroscopic phenomena,and provides decision-making basis for virus prevention and control,rumor control and product promotion by predicting the spread size and threshold.When spreading research of complex network is applied to the recommendation scenario,it focuses on the microscopic perspective and uses the propagation phenomenon to improve the performance of the recommendation algorithm.However,in the research of biological and social spreading mechanisms on complex networks,there is a lack of spreading dynamics models that comprehensively consider population mobility and resource allocation.Besides,the coupled evolution of multiple propagation dynamics also increases the difficulty of research.In the recommendation scenario,it’s difficult for early recommendation algorithms to capture high-order semantic information in the network and recommendation algorithms are faced with data sparsity.How to model and utilize high-order propagation phenomena in recommendation scenarios,and how to effectively integrate social information into recommendation results become the key problems of the recommendation algorithm.To address the above challenges,this thesis first investigates the mechanism of biological propagation on complex networks.Considering the impact of population mobility and resource allocation on disease spread,this thesis constructs metapopulation network model that integrates resource allocation and disease spread.Based on the microscopic Markov chain theory and Monte Carlo simulation,this thesis proves that the resource allocation strategy between cities without structural tendency and considering the severity of the epidemic situation in various places is the most conducive to controlling the disease.Under the optimal resource allocation strategy,the appropriate population movement does not cause large-scale outbreaks of disease.Next,this thesis studies the biological-social co-evolving spreading mechanism on complex networks.We innovatively model the interaction relationship between information diffusion,resource allocation,and virus propagation.Considering the latent state of the virus,we construct SEIS spreading model with information-driven resource allocation.Based on the microscopic Markov chain theory and Monte Carlo simulation,this thesis finds that the diffusion of relevant infonnation inhibits the spread of the disease.Short incubation period will cause first-order jump,which will aggravate the vulnerability of the system.But improving resource efficiency will turn the first-order jump into a continuous one.Finally,this thesis applies complex network propagation to recommendation algorithms,and proposes a recommendation algorithm SRCS based on information propagation on rating and social network.The algorithm builds ratings and social networks through real data,and denoises social networks based on user interest similarity.Then,based on the layer-by-layer convolution of the Graph Neural Network,the information diffusion on the network is realized.Then the high-order propagation phenomenon on the network is modeled,and the user vector representation of the rating and social perspective is learned.In order to integrate social information into the recommendation results,a contrastive learning auxiliary task is designed to achieve dual-view collaborative supervision.Experimental results on real data sets show that the SRCS algorithm has better performance.