CLUSTER NUMBER ESTIMATION PROBLEM IN PARTITION METHO
Санкт-Петербургский государственный электротехнический университет „ЛЭТИ“, кафедра вычислительной техники; аспирант
M. S. Kupriyanov
Saint Petersburg Electrotechnical University “LETI”, Saint Petersburg, 197376, Russian Federation; Head of Prospective Areas
Read the full article
Abstract. The problem of cluster number estimation in analyzed set partition methods is considered. The set of wellknown criteria designed for centroid methods are discussed: partition coefficient, partition entropy, and partition efficiency. Two new criteria are proposed: modified partition coefficient (for centroid methods) and partition quality (for non-centroid method).
Keywords:
clustering, partition methods, fuzzy equivalence relation, solution quality, quality criterion, partition coefficient, partition entropy.
