Researches On Multidimensional Visual Analysis Based On Dimension Subdivision

Multidimensional and high-dimensional data are prevalent in daily life and scientific research.Information visualization of multidimensional data has always been a research hotspot.The complexity of multidimensional data is not only reflected in huge number of dimensions,large data scale,but also in the complex relationship between dimensions.Therefore,the analysis of patterns and associations between dimensions is also of great significance to the study of multidimensional data.Most of the existing methods analyze one dimension as a whole.These methods considered all the correlations between dimensions are at the same granularity.But in the real world,the correlations may be multi-scale.Therefore,this paper will use the method of dimension subdivision to carry out dimensional analysis research on multiscale correlation.In fact,dimension subdivision will increase the number of dimensions,which is not popular in traditional multidimensional data analysis.But multi-level dimensions can also reflect multi-scale data associations,which has certain value for analyzing multi-dimensional data.In this paper,a dimension is subdivided into several sub-dimensions by the proposed dimension subdivision method,then the correlation between dimensions and sub-dimensions is calculated to represent the multi-scale correlations,and the correlation differences before and after subdivision are analyzed.A multi-scale correlation analysis based visualization system is designed.The system provides suitable visualization views,rich interaction and visualization methods to help users analyze data space and dimension space at the same time.In this paper,the force-oriented graph is used as the base view to map the correlation matrix and the overlap matrix as the dimensional view.This system also provides several auxiliary views for the data space,which can be linked to the dimensional view.In addition,this system can save and display multiple dimension views and auxiliary views during historical analysis.Through the experiments of a synthetic data set and two real data sets and the user study,this paper demonstrates the effectiveness of the method and system in multi-scale dimensional analysis.