Research On Visual Analysis Of Ego-centric Networks

Ego-centric network refers to the social network composed of the specific individual and the individuals connected with it.The analysis of ego-centric networks can reveal the laws of the community structure,information dissemination and individual behavior in the social network.Because of the complexity of node characteristics and diverse observation scale in ego-centric networks,single data mining method is difficult to carry on the comprehensive exploration,and the visual analysis method becomes the inevitable choice.The visual analysis of ego-centric networks faces two major problems.One is to use a reasonable layout method to accurately reflect the characteristics of the network,and the second is to propose an effective method for multi-level visual analysis of multivariate egocentric networks.Most of the existing methods are based on academic cooperation network data,but online social network data is more random,lower signal to noise ratio.And such methods are difficult to accurately reflect its characteristics.Therefore,based on the typical online social network – Weibo,this thesis focuses on the above two major problems and proposes a new layout method and a hierarchical visual analysis solution for ego-centric networks.The research results are as following:1.An ego-centric network visualization layout method based on circle detection,Circle Map,is proposed.The method uses the circle-individual two-layer layout structure,using ego-centric network structure information,circle division and node topic distribution as input,and the final layout result is obtained by circle packing positioning,node MDS position calculation,Voronoi diagram generation and contour pruning.In this thesis,the layout results of Circle Map are compared with the traditional layout methods including structure based MDS layout and community division based force-directed layout.The comparison proves that Circle Map method can accurately reflect the social community distribution,structure feature and topic distribution in egocentric network.At the same time,Circle Map can support multi-scale scaling,which brings advantages in the multi-level visual analysis.2.A hierarchical visual analysis method of multivariate ego-centric networks is proposed.Aiming at the multivariate multi-level ego-centric networks,this thesis establishes the analysis model of macroscopic-mesoscopic-microscopic levels.The visualization view and visual mapping are designed in detail at each level,and an interactive visual analysis pipline is constructed.The macroscopic level analyzes the distribution of the nodes and the information flow characteristics implied by the network through the community structure view and the topic view.In the mesoscopic level,the characteristics of each community and their relationship are analyzed by Circle Map view.At the same time,an evolution analysis method of the ego reposting network is given.The microscopic level analyzes the influence of members and their role in information dissemination through the circle structure view,and analyzes how social information flows to the ego through the reposting model analysis.3.Based on the above method,a hierarchical visual analysis system of multivariate egocentric networks is implemented.A real individual in Weibo is chosen to carry out the case study,which prove that analysis tasks proposed in this thesis are well accomplished and the visual analysis design is effective.