Research On Visual Analysis Of High Dimensional Data

High-dimensional data is a kind of common data in life.The analysis for its grouping characteristics and timing characteristics is very challenging.Compared with traditional data analysis methods,visualization and visual analysis technology is intuitive and interactive.Visual mapping allows users to visually perceive the association among dimensions,data distribution,and timing characteristics in high-dimensional data.Reasonable interaction design allows users to explore the data in depth,to help users analyze highdimensional data without any prior knowledge.In this paper,the grouping characteristics and timing characteristics and interaction design in high dimensional data are studied.High-dimensional data and high-dimensional time-series data are visualized by reasonable visualization methods,and the users can explore and analyze the high-dimensional data through effective interactive strategies.The main contents of this paper are as follows:(1)Interactive adaptive parallel coordinate plot.High dimensional data has obvious grouping characteristics.Traditional parallel coordinate plot can show the overall distribution of data and dimension association,but it is difficult to complete the analysis tasks within and between collections.In this paper,the traditional parallel coordinate plot is improved,and an interactive adaptive nested parallel coordinate plot is presented.The multiple sets of high-dimensional data can be visualized,and then they are compared and analyzed.The data order has a great impact on the display results.Therefore,this paper proposes two kinds of dimension rearrangement algorithms to improve the user's perception of high-dimensional data from dimensional relations and inter-group relations.(2)High-dimensional time series data visualization method based on state diagram.High-dimensional time series data has both high-dimensional and time-varying characteristics.Traditional methods such as river map or parallel coordinate map cannot show the dimension correlation and time-varying features in the data well.In this paper,a highdimensional time-series data visualization method based on state diagram is proposed.First,the high-dimensional data is projected onto the two-dimensional plane to obtain the state diagram and the high-dimensional time series data are visualized.Through the distribution of the nodes in the state diagram,the user can know the state distribution and state transition of the data in the time-varying process.(3)Visual interaction design based on multi-view linkage.In order to analyze highdimensional time series data in a multidimensional and multi-angle way,a visual interaction design based on multi-view linkage is proposed in this paper.The method uses the state diagram to produce the overall perception of the data;the parallel coordinate map displays the original data;the scatter matrix explores the relationship between each dimensions;the thematic river and thermogram show the process of data of different dimensions changing with time from different angles.The design of a variety of interactive means and multi-view linkage between the data helps users to explore the data deeply.