The Comparison of Fuzzy Clustering Methods for Symbolic Interval-valued Data

Interval-valued data can find their practical applications in such situations as recording monthlyinterval temperatures at meteorological stations, daily interval stock prices, etc. The primary objectiveof the presented paper is to compare three different methods of fuzzy clustering for interval-valuedsymbolic data, i.e.: fuzzy c-means clustering, adaptive fuzzy c-means clustering and fuzzy k-meansclustering with fuzzy spectral clustering. Fuzzy spectral clustering combines both spectral and fuzzyapproaches in order to obtain better results (in terms of Rand index for fuzzy clustering). The conduc-ted simulation studies with artificial and real data sets confirm both higher usefulness and more stableresults of fuzzy spectral clustering method, as compared to other existing fuzzy clustering methodsfor symbolic interval-valued data, when dealing with data featuring different cluster structures, noisyvariables and/or outliers.