APPLICATION OF MACHINE LEARNING METHODS FOR STATISTICAL ANALYSIS AND FORECASTING OF THE E-SPORTS INDUSTRY

Abstract

The article considers the dynamics and behavior of the e-sports industry at the global level and the state of e-sports as an industry in Ukraine. The main achievements of the e-sports sphere of Ukraine are determined. The statistical analysis of the income of the e-sports industry, the total audience of e-sports games, regular and average spectators of competitions on the basis of the analysis of variation, fashion, indicators of asymmetry and excess of distribution is carried out. To achieve the objectives of the study, the method of exponential smoothing and the method of reference vectors were used. SVM is a machine learning method used to solve classification and regression problems. As for the classical regression model, the basis of the approach is to find the function of fitting empirical data. The chosen methods allowed to prepare data for analysis and to build regression SVM-models with a kernel on the basis of radial-basis functions. The built models for the income of e-sports and ordinary e-sports spectators are of the epsilon-SVM type, and for the global audience of e-sports and regular e-sports viewers - nu-SVM. The adequacy of the constructed models is proved on the basis of the analysis of model residues. Input indicators are predicted. It is determined that by 2025 the income from e-sports activities is expected to grow steadily, which means the constant development and improvement of the infrastructure related to e-sports. The importance and necessity of state support for the development of e-sports at all levels: from the organization of tournament venues to regional, school, amateur tournaments. The obtained results can be used by the Federation of e-sports of Ukraine, e-sports organizations, researchers to substantiate the need for the development of e-sports in Ukraine.