Research On Network Public Opinion Data Visualization Technology

With the development of social media,network public opinion data has grown dramatically,then the effective network public opinion data visual analysis plays an important role in dredging paradox and developing strategies.Due to the wide range of data acquisition and the diversity of access channels,the network public opinion data presents massive characteristics,multi-dimensional characteristics and spatio-temporal characteristics.Based on these characteristics,this thesis has done a lot of research work on massive level data visualization and multi-dimensional spatio-temporal data visualization.The treemap is one of the most powerful spatial filling techniques for hierarchical data visualization.However,as the amount of data increases,the nesting levels of treemap become too much,and the layout algorithm brings great visual confusion.Interactive exploration is an effective way to alleviate visual clutter in treemap.It reduces the number of nodes visualized in the view by increasing the interaction.However,the existing interactive exploration methods have some drawbacks,such as the translation and scaling on the two-dimensional plane provide the limited context information,and the view query methods require prior knowledge.Therefore,this thesis proposes the minimum description length principle based on user interaction for unbalanced weighting(UMDL)method.When the focus is obtained,the method controls the nodes in the focus area to be directly displayed or displayed by the aggregation node,and always provide high quality views during the interaction.In addition,the method has no conflict with existing interactive exploration methods,and the method can be selectively combined according to different scenarios to achieve better results.It is a common method to visualize multidimensional spatiotemporal data by obtaining metrics in each dimension by tensor decomposition and visualizing each dimension separately.However,as the amount of data increases,it is difficult to provide a real-time view.And as the data collection time increases,the amount of data in the time dimension increases,and the time span becomes wider,improving the flexibility of the time dimension becomes critical.However,existing visual analysis frameworks are difficult to meet this requirement.Therefore,this thesis proposes a multi-resolution tensor decomposition visual analysis framework.The framework builds multi-resolution tensors along the time dimension using the hierarchical structure inherent in time granularity.In addition,this thesis accelerates the view update in data exploration with a multi-level storage structure that accelerates the construction of tensors.Based on the above research,this thesis implements the network public opinion data visual analysis system.The system implements functions such as data acquisition,data filtering,data storage,and data visualization.In data visual analysis,the system pro-vides multi-view collaboration to provide different perspectives for data analysis,helping data analysts to better compare,analyze and verify.