Journal of Classification 4:7-31 (1987)
Journal of
Classification
©1987Springer-Verlag New YorkInc.
Additive Clustering and Qualitative Factor Analysis
Methods for Similarity Matrices
B. G. Mirkin
Central Economics-Mathematics Institute of USSR's Academy
Abstract: We review methods of qualitative factor analysis (QFA) developed
by the author and his collaborators over the last decade and discuss the use
of QFA methods for the additive clustering problem. The QFA method
includes, first, finding a square Boolean matrix in a fixed set of Boolean
matrices with "simple structures" to approximate a given similarity matrix,
and, second, repeating this process again and again using residual similarity
matrices. We present convergence properties for three versions of the
method, provide "cluster" interpretations for results obtained from the algo-
rithms, and give formulas for the evaluation of "factor shares" of the initial
similarities variance.
Keywords: Additive Clustering; Qualitative factor analysis; Macrostructures;
Average compactness.
Introduction
A model and corresponding algorithms for the additive clustering of
similarity matrices (ADCLUS) have been described by Shepard and Arabie
(1979), Arabie and Carroll (1980), and Arabie, Carroll, DeSarbo, and Wind
(1981). A three way generalization of this model (INDCLUS) and a
corresponding method have also been described by Carroll and Arabic
(1983). The additive clustering model assumes an approximate representa-
tion of the initial similarity matrix in the form of a weighted sum of Boolean
(0, 1) matrices with simple structures that correspond to dusters. A similar
idea motivates the methods of qualitative factor analysis elaborated by the
I am indebted to Professor P. Arabie and the referees for valuable comments and editing
of the text.
Author's Address: Boris G. Mirkin, Central Economics-Mathematics Institute, Krasikova
str. 32, Moscow 117418, U.S.S.R.