Analysis of clustering algorithms for use in the universal data processing system

Abstract

The article offers criteria that allow you to select algorithms that are appropriate for application in a specific subject area, namely the algorithmic complexity of the algorithm, the accuracy of the object’s belonging to the cluster, the attribution of the object to one or another cluster at the boundary of two clusters, the ability to build from clusters fuzzy logic rules. Also described is the architecture of the created system, which clusters the data with different algorithms to use from the desired area. The algorithms were tested to solve the problem of data clustering to predict the degree of server energy efficiency at certain values involved in the processor frequency and number of cores calculations. The use of algorithms allowed us to analyze and look at the features of each, to determine the fuzzy C-average as the most appropriate for this task.