The Influence Of Sign Division On Signed Networks

With the rapid development of information,a lot of interactive information emerges on the Internet,which naturally brings a lot of score information.These interactive information show the tendency of emotion,so it is convenient for signed networks to effectively add these emotional information to the sub-sequent research and analysis.The positive and negative sides of the signed networks indicate that there is a positive and negative relationship between two nodes.In the real world there are many complex network relations,especially in the field of information,biology and sociology,using the edge of the symbol attribute to analyze network topology structure and understand the network function and predict dynamics behavior has important theoretical significance,this attitude to relationship prediction,prediction,personalized recommendation and user characteristics analysis and association analysis has important application value.The research on any signed networks is divided into two stages.The first stage is from the original data to the construction of network.The second part is the follow-up research on the constructed network.Obviously,the work in the first phase is the basis of the work in the second phase.When we build the network in different ways,the same data set studying the same problem may produce different results Based on the above considerations,the focus of this paper is to divide the scoring information into different situations,and at the same time,some indicators are used to explore whether different classifications will have an impact on the network structure.The main research achievements are as follows:By analyzing the data structure of the real data sets of Netflix,Movieles,Book and Douban,the middle score was determined,and the middle score of Netflix,Movieles and Book was 3 points and the middle score of Douban was 4 points.By using the indexes of positive and negative correlation degree,signed networks stability and so on,the influence of different partition of intermediate score on signed networks is analyzed.Through the research of empirical network,it is found that there is little difference in different forms of division.In addition to the empirical study of signed networks under different partition conditions,it is still a question that we should pay attention to whether the different network types and the proportion of intermediate score will have an impact on our research results.Therefore,relevant index analysis is carried out for the artificial network,in order to understand whether the artificial network is also suitable for the conclusions drawn in the empirical network,and to exclude contingency.For this reason,taking small-world network as the basic network structure,considering the influence of the intermediate score of different proportions on the signed networks,many signed networks were randomly constructed.Meanwhile,the signed networks with different proportions of intermediate score was studied by using the relevant indexes on the signed networks Through the research on the artificial network,it is found that the processing method of ignoring the intermediate score is unreasonable,because it will reduce the structural information in the network,thus affecting the subsequent network research,which is crucial for the later research on the signed networks...