Abstract:
In the ever-evolving world of financial markets, accurately predicting stock prices is a perpetual challenge. In our research, we've taken a two-pronged approach, combining traditional time series analysis with cutting-edge machine learning techniques to forecast stock prices. We've applied the power of various machine-learning algorithms, including LSTM, RE, SVR, XGBoost, ARIMAX, GRU, BİLSTM, CNN, RHN, and TCN. But what sets our research apart is our creation of twelve ensemble models, each strategically blending diferent algorithms. Models like LSTM-RE, LSTM-SVR, and LSTM-XGBoost are designed to work together, capitalizing on their strengths and fixing their weaknesses to enhance prediction accuracy. A critical aspect of our research is feature engineering, where we carefully select specific data attributes. We've integrated four key features derived from historical stock price data: daily return, 5-day moving average, 10-day moving average, and 10-day moving standard deviation. These features capture vital information about stock price momentum, trends, and volatility. Our initial findings are promising, indicating that this combination of ensemble models and thoughtful feature engineering can boost the precision of stock price predictions. This synergy between advanced machine learning methods and valuable feature engineering allows for more accurate forecasts in the complex realm of finance. One noteworthy discovery is the remarkable performance of the LSTM-ARIMAX model, which stands out due to its proficiency in predicting stock values. This model combines LSTM for identifying short-term patterns and ARIMAX's strength in considering external influences and long-term trends. Beyond theory, our research explores practical applications. It digs deep into how both experienced and beginner investors can leverage advanced prediction models to refine their investment strategies. Additionally, it opens up avenues for future research, particularly in real-time stock price prediction and emerging machine-learning approaches. Keywords: finance, ensemble models, feature engineering, LSTM, ARIMAX,
