Anisotropic parallel coordinates with adjustment based on distribution features

To measure the confidence degree of correlation between data dimensions in multidimensional data, we present a visualization method named anisotropic parallel coordinates. The method introduces distribution features of data into classical parallel coordinates scheme. The method first divides the data in each dimension into segments and obtains the frequency of data in each segment. The histogram is adopted to express the distribution of data in each dimension. The coordinate axis of each dimension is adjusted according to the corresponding distribution features. The principle of the adjustment is to amplify the occupation in the axis for the data segment with biggish frequency, while compacting the segment with lesser frequency. The adjustment can improve the capability of expressing the correlativity between the adjacent dimensions effectively in the final visualization result. The experimental results prove the method presented in the paper can achieve more effective expression to the correlativity between the adjacent dimension data. The improved effect can enhance the efficiency of the visual interaction and the visual analysis for the multidimensional data.