Isodata, a Novel Method of Data Analysis and Pattern Classification

摘要

ISODATA, a novel method of data analysis and pattern classification, is described in verbal and pictorial terms, in terms of a two-dimensional example, and by giving the mathematical calculations that the method uses. The technique clusters many-variable data around points in the data's original high-dimensional space and by doing so provides a useful description of the data. A brief summary of results from analyzing alphanumeric, gaussian, sociological and meteorological data is given. In the appendix, generalizations of the existing technique to clustering around lines and planes are discussed and a tentative algorithm for clustering around lines is given. (Author)