Інформаційна технологія прогнозування курсу криптовалют на основі комплексної інженерії ознак

Abstract

This article is devoted to the development of information technology for predicting the rate of cryptocurrencies based on complex feature engineering. The peculiarity of this technology lies in a systematic approach to the ion of features. An analysis of groups of external and internal factors of the potential impact on the cryptocurrency market was carried out. An analysis of the features that characterize changes in the cryptocurrency rate showed that apart the basic features available on many cryptocurrency exchanges. The most important for further prediction of cryptocurrency rates are secondary features which are derived the basic ones by applying different mathematical operations and/or algorithmic transformations to them. The analysis of a large number of sources showed that cryptocurrencies have a number of characteristics, which caused their great popularity and also should be taken into account during the formation and ion of features. The work systematizes the key to such characteristics and suggests how to formalize them as features. It is suggested to formalize the features by a systematic approach, according to the postulates of technical cybernetics, which says that any research object can be represented in the form of a black box (BB), which is in contact with the external environment at nine points, which in a rich case can be a multiplicity of features or variables. A general mathematical model of the formation of these factors is proposed, which consists in generating a large number of secondary features on the basis of simple mathematical, algorithmic and statistical transformations with a further ion of the most relevant of them. The technology involves the synthesis of new secondary features on the basis of other secondary features, with some exceptions, which are formalized in the form of a system of rules. This will reduce the overlearning of the model and improve its overall ability. In order to explain the validity of the developed technology, the application of its use on the Bitcoin exchange rate data for 2020-2021 was examined. The conducted research and computer experiments showed the efficiency and usefulness of the proposed technology.