Temporal Network Analysis And Its Application In The Character Relationship Network

As extensions of static networks,temporal networks can effectively cover time information so as to model networks that change over time.Many real-world systems have time attributes,such as human proximity networks,citation networks,economic networks,collaboration networks and transportation networks.Therefore,temporal networks have a wide range of applications.This thesis constructs temporal networks of character relationship networks and analyzes networks using temporal methods.Based on temporal motifs,this paper studies the evolution of the characters’ relationship,the development of stories,the evolution of community structures and the topology of networks.We aim to do deep researches and explore applications in temporal network area.It mainly includes the following work:(1)A model of temporal networks based on two dimensions,edge principle and time interval,is constructed.The time interval of networks is progressively narrowed,which builds a good foundation for latter contrast analysis.(2)A star network algorithm based on temporal motifs is proposed and topology structures of weighted temporal networks are analyzed.Based on measures including network density,node degree and average path,the characteristics of networks’ evolution are analyzed.Based on strength and information indices,the importance and changes of characters are analyzed.Based on s-core and optimized s-core,the core groups and changes are analyzed.(3)Motif structures in character networks are excavated from both static and dynamic perspectives,and temporal motifs are applied to character relationship networks for the first time.We also apply specific characters’ meanings to motifs in order to analyze the relationships between characters in detail.The static motifs are used to analyze superfamilies of networks and dominant motifs,while the dynamic motifs are used to analyze relationship changes of characters and consistency of the contents in the novel.(4)A community mining algorithm based on temporal motifs is proposed.Adding temporal motifs to the traditional community mining algorithm can quickly deal with the problem of temporal networks’ community divisions.On the premise of better partitioning,it fully reflects the time-varying characteristics of the temporal networks.In addition,based on the results of community divisions,the characteristics of characters are analyzed in detail.