Digital competitive advantage of the exchange adviser based on the deep learning model «random forest regression»

Abstract

The article deals with the theoretical foundations of the use of algorithmic trading technologies. The relevance of the study lies in the fact that in modern conditions there are more and more exchange robots, their algorithms are being improved and the technical capabilities of their application are expanding, moreover, increasing the competitiveness of exchange bots is becoming an acute problem, from the standpoint of increasing forecasting accuracy, which ensures their digital competitive advantage. The novelty lies in the fact that the study proposed a developed stock advisor based on the Random Forest Deep Learning model "Random Forest Regression". The purpose of the study is to develop an algorithm for predicting the closing price of a SiU3 dollar futures contract on the Moscow Exchange. To achieve the goal, the following tasks were set and solved: to explore the theoretical foundations for the use of ensembles of decision trees and develop an exchange advisor based on the Random Forest Regression Deep Learning model, as well as to form a VaR model for assessing fi nancial risk, in addition, to study competition in the fi eld of algorithmic trading in the stock market. The practical significance is that the results of the study can be used in practice in exchange trading. As a result, a high accuracy of the forecast was obtained, since the level of the average absolute error of the model does not exceed the value of 49.01 rubles, and the level of the maximum absolute error does not exceed the value of 162.52 rubles, while MSE (R2) is equal to –0.2573 for the option in which the share of the test sample in the dataset was 0.20.