Research On The Optimization And Application Of Small-world Network Model

Small-world Theory is also called the Theory of Six Degrees of Separation,which means that the association interval between any two individuals in a real society is statistically no more than 5 people.It is found that many complex networks have similar structures with Six Degrees of Separation,for example,the network structure of business information in economic activities and the food chain network structure in biological systems,etc.The high efficiency and convenience brought by the short distance is the application value of Small-world Theory.In this thesis,the application of Small-world Theory includes two aspects,one is optimizing the small world of e-commerce information dissemination network in improving dissemination efficiency and maximizing commercial value.The other is combining Small-World with artificial neural network,and realizing the intelligent learning of public opinion with it,in order to achieve high efficiency and accuracy in pre-warning.The research work of the thesis includes the following three aspects.(1)Key nodes mining in e-commerce information dissemination network.The mining of key nodes is the basis for optimizing the networks model.In this thesis,the key nodes are divided into central key node that functions as a community cluster and the connectivity key node that connects different communities.The mining of central key nodes bases on the traditional PageRank algorithm.The PR value distribution of traditional PageRank has been improved,and the influence distribution factor is the focus from other non-key nodes.A page correlation based on EC-PageRank algorithm has been designed.The effect of the mining algorithm is verified by real data.The mining of connectivity key nodes bases on overlapping community mining algorithms,and the key nodes are taken in overlapping community areas.In order to improve its efficiency and quality,a mining algorithm for overlapping community in e-commerce network based on central node has been designed.The algorithm uses the central key node as the initial seed,and improves the community fitness function with the frequency of actual interaction between users,finally adopts a reasonable merger strategy for highly overlapping communities.The effectiveness of community mining algorithms has been verified on simulation networks and real networks respectively.(2)The Small-world optimization of e-commerce information dissemination network model.The optimization scheme aims at shorter network average path length and higher clustering coefficient.The specific implementation of the scheme includes the selection of reconnected operation nodes and the selection of reconnected operation edges.The thesis selects the key node as the computing node of the reconnected edge adding or deleting,and it deletes the neighboring edge with the smallest decrease in the network average information transmission efficiency from the node.When the adjacent edges are deleted,the edges are reconnected based on the rule that maximize the incensement of average information transmission efficiency.The optimization scheme improves the small-world character of the information dissemination network model,and further improves the efficiency of information dissemination,and the improvement of efficiency is verified by experiments.(3)Research on the early warning system of microblog public opinion.The thesis firstly integrates the existing public opinion early warning index system,considers two aspects of strength and popularity,designs two first-level indicators,and gives nine quantifiable secondary indicators aiming at reducing the redundancy between the indicators.Taking the designed index system as input,this thesis designs a small world neural network public opinion early warning system based on hybrid optimization algorithm.The hybrid optimization algorithm is based on genetic algorithm and combines the advantages of strong global search ability of genetic algorithm and strong local search ability of simulated annealing algorithm.It performs simulated annealing operation on the middle group generated by genetic algorithm in iterative process to effectively avoid falling into local optimum.