Abstract
The article describes the optimization problem solving for multidimensional and bulks Big Data (data with more than 10 characteristics and \(10^8\) observations or higher), as well as machine-generated data of unlimited volume. It is difficult to analyze and visualize data of such volume and complexity using traditional methods. In contrast (and in addition) to machine learning methods widely used in Big Data analysis, it is proposed to use stochastic methods of data sets’ coding and approximation using Kolmogorov-Shannon metric nets, which are optimal for the entropy of the code. While adapting these methods, new methods are proposed for metrics construction for characteristics with nominal and ordinal scales.