CLUSTERIZATION OF DATA ARRAYS BASED ON COMBINED OPTIMIZATION OF DISTRIBUTION DENSITY FUNCTIONS AND THE EVOLUTIONARY METHOD OF CAT SWARM

Abstract

Context. The task of clustering arrays of observations of an arbitrary nature is an integral part of Data Mining, and in the more general case of Data Science, a huge number of approaches have been proposed for its solution, which differ from each other both in a priori assumptions regarding the physical nature of the data and the problem, and in the mathematical apparatus. From a computational point of view, the clustering problem turns into a problem of finding local extrema of a multiextremal function of the vector density argument using gradient procedures that are repeatedly launched from different points of the initial data array. It is possible to speed up the process of searching for these extremes by using the ideas of evolutionary optimization, which includes algorithms inspired by nature, swarm algorithms, population algorithms, etc. Objective. The purpose of the work is to introduce a data clustering procedure based on the peaks of the data distribution density and the evolutionary method of cat swarms, that combines the main advantages of methods for working with data in conditions of overlapping classes, is characterized by high-quality clustering, high speed and accuracy of the obtained results. Method. The method for clustering data arrays based on the combined optimization of distribution density functions and the evolutionary method of cat swarms was proposed. The advantage of the proposed approach is to reduce the time for solving optimization problems in conditions where clusters are overlap. Results. The results of the experiments confirm the effectiveness of the proposed approach in clustering problems under the condition of classes that overlap and allow us to recommend the proposed method for use in practice to solve problems of automatic clustering big data. Conclusions. The method for clustering data arrays based on the combined optimization of distribution density functions and the evolutionary method of cat swarm was proposed. The advantage of the proposed approach is to reduce the time for solving optimization problems in conditions where clusters are overlap. The method is quite simple from the numerical implementation and is not critical for choosing an optimization procedure. The experimental results confirm the effectiveness of the proposed approach in clustering problems under conditions of overlapping clusters.