Stock Price Forecasting Using Long Short-Term Memory Networks: A Comprehensive Analysis of Financial Time-Series Data

In this study, we employ Long Short-Term Memory (LSTM) networks to predict stock prices using historical data from Tesla Inc. spanning 10 years. The research emphasizes the importance of LSTM's capability to model complex temporal dependencies in financial time-series data, outperforming traditional statistical methods. Various feature combinations and time steps are tested, identifying a 30-day window as the optimal setup. Key evaluation metrics, including Mean Squared Error (MSE), training loss, and validation loss, are utilized to assess model performance. The results show that LSTM networks are particularly effective for short-term stock price predictions, while longer-term forecasts experience decreasing accuracy. This work contributes to the growing body of knowledge on deep learning applications in financial markets, offering practical insights for investors and financial institutions