Classical and recent approaches in cluster analysis

The paper focuses on the development of selected approaches in cluster analysis. There are recently proposed similarity measures for objects characterized by nominal variables, development of algorithms for k-clustering and development of methods for clustering large data files and categorical data. As concerns algorithms for k-clustering, attention is paid to take into account the uncertainty in classifying objects into clusters, namely FCM (fuzzy k-means), PCM, FPCM, RCM, RFCM and RFPCM algorithms. For large data files, algorithms CURE, ROCK, CLARA, CLARANS and BIRCH are included, for categorical data clustering there are COOLCAT and ROCK algorithms. Two-step cluster analysis to cluster large data sets with variables of different types is mentioned