Visual clustering methods with feature displayed function for self-organizing

To improve the intelligibility and visibility of clustering, through digging spatial informations which hide in sample vectors and advancing the analytical method of significant feature item and the class-feature standard deviation, showing the chiefly factor engenderd clustering and each feature item´s contribution rate to clustering. This scheme realizes dynamic visualization display clustering procedures, optimum cluster and the conclusion of analyzing feature item intuitively, which supplies assistances and offers clues to recognize the work process and arithmetic of neural network. The emulation experiments show that this scheme has grate value of theoretical research and engineering application.