<![CDATA[Cryptocurrencies have emerged as one of the most popular digital assets, characterized by high volatility, which presents a significant challenge in forecasting their price movements accurately. This study aims to implement the Long Short-Term Memory (LSTM) algorithm to predict the prices of selected cryptocurrencies, including Bitcoin (BTC), Binance Coin (BNB), Ethereum (ETH), Dogecoin (DOGE), Solana (SOL), and Shiba Inu (SHIB). The LSTM model is trained using the Adam optimizer and employs early stopping to mitigate overfitting. Model performance is evaluated using Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R²). The results indicate that the LSTM model achieves strong predictive accuracy for relatively low-volatility assets such as Dogecoin and Solana, with R² scores of 0.9795 and 0.9523, respectively. In contrast, its performance declines when applied to highly volatile assets like Bitcoin and Binance Coin. The findings also suggest that LSTM performs best in short-to-medium-term forecasts (7 to 30 days), but shows limitations in long-term predictions. This study contributes to the field by demonstrating the applicability of LSTM in financial forecasting and highlighting its strengths and constraints across different volatility profiles. Practically, the findings can assist traders and financial analysts in making data-driven decisions by applying LSTM models for more reliable short-term predictions, while emphasizing the need to integrate external market factors to enhance long-term forecast accuracy.]]>
Copyrights © 2025
