Study on Feature Extraction Method Based on Parallel Coordinate Plots

Abstract

A novel feature extraction method based on parallel coordinate plots was presented. Observing the parallel coordinate plots, discovered that using the distance of one point to others on one dimensionality to measurement the classify performance of the variable, can express the fact classify performance more impersonally. The Euclidean distance or module matrix and the relative distance matrix were given. And the distance ratio of every sample point to other sorts and it to its own sort has more classify information. We achieved better performance when experiment on data which has poor statistical performance.