Spectral clustering that was proposed by Ng, Jordan and Weiss, is not in fact a new clustering method, but rather a new way to prepare data set for clustering method. This method uses the idea of spectral decomposition. The main aim of the paper is to present a possibility of application spectral clustering when dealing symbolic data, with a special focus on different distance measures that can be applied for this kind of data. In experiment studies artificial data sets with known cluster structure were obtained with application of clusterSim and mlbench packages of R software. Each data set was clustered 40 times with application of each distance measure applied. Received results were compared with known cluster structure with application of adjusted Rand index.
JavaScript is turned off in your web browser. Turn it on to take full advantage of this site, then refresh the page.