УПРАВЛІННЯ ПОРТФЕЛЕМ ЦІННИХ ПАПЕРІВ З ВИКОРИСТАННЯМ МЕТОДІВ ІНТЕЛЕКТУАЛЬНОГО АНАЛІЗУ ДАНИХ У АНАЛІЗІ ЧАСОВИХ РЯДІВ

This study explores the application of Data Mining techniques, specifically deep neural networks (DNN) and recurrent neural networks (RNN), for optimizing stock portfolios. Using time series data, we compare the performance of DNN and RNN models in predicting stock prices and constructing optimal portfolios. Key evaluation metrics demonstrate that the RNN model's forecasts yield a portfolio with income and risk metrics that closely match actual values, outperforming the DNN model. Furthermore, the RNN model's portfolio weights show a stronger alignment with actual distributions, indicating superior predictive accuracy in asset allocation. This study concludes that RNN, with their inherent capability for processing sequential data, are particularly well-suited for time series forecasting in financial applications.