Design And Implemention Of Visual Analytical System Based On Social Data

The rise of social networks in the time of Information Revolution has completely changed the way people communicate with a large amount of social data producing.Traditional methods for large-scale,high-dimensional social data visualization are difficult to identify human social behavior patterns.Through filtering selection,clustering and other technologies can solve the occlusion,confusion,interaction and other issues raised from large-scale social data visualization to a certain extent,but relevance of different social type data is neglected,such as association between online friendship data and social check-in data.Such association cannot be mined by simple selection or clustering techniques.This thesis combines social friendship data and social check-in data to explore the social behavior patterns of social users,and realizes the social portrait system to visually analyze multiple social big data.The main work of this thesis is as follows:(1)The social network is described by a graph model,and the community detection algorithm is applied to calculate the characteristics of the social network.Features of the check-in data are extracted based on the social community structure of the social users.The combined feature calculation of the association between the two types of social data not only makes the social users‘ community explainable by POI type of offline activities,but also selects social user with high community interpretability by the way constructing the RCVF parameter.(2)For social users with high community interpretability,unsupervised clustering algorithm is used to classify different social patterns and reduce the clustering results to 2 D dimension.The clustering method not only can get the result of social type division,but also provides a reasonable ground for the next sampling.(3)Perform reasonable social pattern sampling.Visual design in this thesis is used to display the social information of the same pattern of sampled social users,and then a social user can be selected in a certain social type to display its multi-faceted social details.Hierarchical visual analysis of social data is achieved from overview to a certain type of social pattern,ultimately to personal social user.(4)A reasonable visual design for the extracted social features.by coding social features of various social types and various aspects(such as spatio-temporal features)into only one visual design in a limited space,making the social features which are necessary to present can be expressed reasonably and efficiently.The data analysts can conduct more effectively combined analysis of multiple types of social data,while comprehensively understanding the details of social data.(5)A system combined visual feature abstraction and social pattern recognition based on multiple types of social data—Social Portrait is designed and implemented.Case studies based on real multi-type social data confirm the utility for social big data and the reliability of visual analysis.