Visual Analysis of Correlation in Multidimensional Data Based on Dimension Projection Technique

Aiming at the requirement of correlation analysis in high dimensional multivariate data,a correlation measure KNN-Pearson based on KNN and Pearson coefficients is proposed firstly.This method can quantitatively express the contribution degree of a dimension to clustering.The method uses the data density of a dimension value as the calculation element,and calculates the correlation between the dimensions by the Pearson correlation coefficient.Then,a visual analysis method of correlation based on dimension projection is proposed.The method uses the multidimensional scale technique to carry out the projection of the dimension,and shows the correlation between the dimensions by using the projection scatter plot and the matrix thermal graph.The distribution and clustering characteristics of the data are displayed by the data projection matrix and parallel coordinates,allowing the subspaces of interest to be constructed by the dimension selection,and the data can be analyzed interactively in the subspace.The visual analysis method of correlation is applied to the field of food safety,and a visual analysis system of pesticide residue detection data is designed and implemented.Through the interaction of data screening,dimension selection,scale scaling and multi-view linkage,the correlation analysis of pesticide detection in multi-regional agricultural products was realized,and the pattern of pesticide application in agricultural areas was found.The effectiveness of the method is proved by user experience and evaluation.