Usage of Deep Neural Networks for Stock Price Prediction Using Technical Indicators

This thesis explores the potential of deep neural networks, specifically RNN, LSTM, and GRU models, for stock price prediction based on technical indicators. The study focuses on bridging theoretical knowledge of technical analysis with modern deep learning methods. The theoretical part provides an overview of financial market technical analysis, its tools, graphical methods, and main technical indicators. It also includes a detailed explanation of the principles of neural networks, their structures, learning processes, and optimization. The practical section involves data preprocessing, implementation of selected models, and their evaluation on historical data. The models were tested using different performance metrics and their results were compared. Additionally, two investment strategies were developed and tested: "buy and hold" and a strategy based on local maxima and minima. The results demonstrated that LSTM and GRU models achieve higher accuracy and stability compared to traditional RNN, particularly in predicting long-term trends. Visualization of predictions and decision points provided valuable insights for optimizing investment strategies. This thesis contributes to the understanding of applying neural networks in financial time series analysis and their potential in practical trading.