Research On Multi-label Seed Node Selection In Social Network Environments

With the popularity of social network,it has been integrated into people's lives.So it has been attracted many scholars' attention.One of the areas of social network is multi-label classification.Now,most scholars only focus their research on how to improve the accuracy of classification,but neglect the problem of system overhead which caused by the expansion of network size.The larger the social network,the more complex the data structure.And it will increase the time and memory which to get the node information.In addition,it will increase the burden on the computer and reduce the operating efficiency.In order to solve this problem,this dissertation proposes a novel seed node multi-label classification method.This method not only greatly reduces the system overhead but also improves the accuracy of multi-label classification.The main work of this dissertation is as follows:(1)This dissertation introduces the background and research on multi-label classification in social network environments,and analyzes the research status and use defects of traditional multi-label classification model and relational classification model.In addition,this dissertation also introduces some related concepts of multi-objective genetic algorithm,and some classical multi-objective genetic algorithm models,with the parameter and application conditions.(2)In order to reduce the system overhead which caused by the acquisition of node information,based on NSGA2 algorithm,this dissertation proposes a multi-label seed node selection algorithm(called NAMESEA).Compared with other algorithms,our proposed NAMESEA algorithm not only saves time,reduces memory occupancy,but also improves the accuracy of classification.(3)In order to provide a better personalized recommendation for users,we should improve the prediction accuracy.Based on the above algorithm,this dissertation extends the objective function.We import a new evaluation function which embeds the relational classification model into multi-objective genetic algorithm.And an improved algorithm(called the MOSS algorithm)is proposed.It not only saved system overhead,but also greatly improved the accuracy of multi-label classification.