Research on ISODATA Techniques. (Interactive Self-Organizing Data Analysis Technique A)

摘要

ISODATA and the Singleton-Kautz algorithm, techniques for clustering multivariate data, are described and some theoretical and experimental results are given. Other techniques briefly described include: PROMENADE, an interactive graphic computer-based system for the analysis of multivariate data (PROMENADE eventually allowed us to control ISODATA interactively and to provide plots and graphs that assist in interpreting data);the Rosen-Hall discrimination algorithm, a discrimination procedure that uses a modification of the ISODATA algorithm, in combination with categories; ISODATA-Lines, a clustering technique for clustering obejects around line segments rather than around points; a technique for permuting the rows and columns of a matrix in order to place either low values or high values near the main diagonal; and a program for positioning cluster centers in a two-dimensional plot in a way that approximates intercluster distance. (Author)