Удосконалений метод кластеризації станів комп’ютерної техніки K-MEANS

Abstract

The work is dedicated to improving the method of clustering classes of computer equipment K-MEANS to improve the quality partition of such states. The object of research is the process of clustering state of computer equipment. Subject of research – methods of cluster analysis states of computer equipment. Relevance of these studies is conditioned by the rapid scientific and technical progress, in which significantly increased the number of computer equipment, which is used in various fields, and therefore increases the likelihood of situations specific to this equipment, given the diversity of functions that it performs. Thus, depending on the state of computer equipment taken various administrative decisions regarding its further functioning. So important is the development or improvement of methods of cluster analysis state of computer equipment that will determine the decision on its further functioning. In the analysis of decomposition of objects that can be used for solving the problem of clustering state of computer equipment, it was determined that such methods must be clear, non-hierarchical and scalable, expressed by characteristics inherent in the clustering method K-MEANS. When tested method known modifications appointed on the possibility of their use in the analysis of the state of computer equipment was found insufficient accuracy of classification of this state to a cluster through a random selection of initial centers. Identified deficiencies have been corrected by determining the initial cluster centers based on the values of potentials, as well as the allocation of a separate cluster of conditions that could be mistakenly attributed to the cluster due to tolerances values of parameters and characteristics of these states, thus improving the quality of the partition of the states of computer equipment an average of 7%.