DPS Clustering: New Results

The results presented in this paper are obtained as part of the continued development and research clustering algorithms based on discrete mathematical analysis. article briefly describes theory Discrete Perfect Sets (DPS-sets) that is basis for construction DPS-clustering algorithms. main task previously constructed DPS-algorithms to search clusters multidimensional arrays with noise. have two stages: first stage recognition maximum perfect set a given density level from initial array, second partitioning result into connected components, which considered be clusters. Study qualities showed that, number situations stage, does not include all practical sense. In components can lead unnecessarily small Simple variation parameters allow eliminating these drawbacks. present devoted their new versions, more free