Применение методов машинного обучения для анализа показателей на финансовых рынках

Abstract

The results of research in the field of application of various machine learning (ML) methods for the analysis and forecasting of financial time series are presented. The purpose of the study is to assess the suitability of ML algorithms for solving the problem of analyzing and predicting the state of financial markets. The source data was public data on Apple shares for the period from May 2002 to June 2020. As part of the study, machine learning algorithms were developed using stochastic analysis and data was prepared with subsequent preprocessing for better learning algorithms, as shown in the example of matrix formation for the boosting algorithm. The area of forecasting financial time series was studied and algorithms of boosting, fully connected neural network, recurrent neural network, neural networks with long short-term memory, convolutional neural network were applied. Based on the trained algorithms, a comparative analysis was carried out on a sample of financial instruments from one sector