Modeling Short-Term Bitcoin Price Dynamics Using Long Short-Term Memory Networks

A Long Short-Term Memory (LSTM) neural network trained on daily BTC/USDT data from the Binance exchange is used in this study to investigate short-term Bitcoin price dynamics. Instead of relying on linear forecasting assumptions, the model is designed to capture temporal dependencies and momentum patterns to address the nonlinear and noise-dominated nature of cryptocurrency markets. A strictly out-of-sample framework is employed to evaluate the prediction task, which is defined as a univariate regression problem with the daily high price as the target variable. According to empirical findings, the LSTM model demonstrates strong statistical performance despite significant market volatility, with an RMSE of USD 3,619.74, an MAE of USD 2,989.73, a MAPE of 2.82%, and an R² of 0.90. Forecasts for the next five days reveal a consistent short-term bearish trend, with broad prediction intervals of approximately USD 6,000, indicating considerable uncertainty and expected prices declining from USD 88,425.62 to USD 85,497.88. The results suggest that LSTM models can extract meaningful trend and regime information, making them suitable as risk-aware decision-support tools rather than deterministic forecasting systems, even though precise short-term price-level prediction remains constrained.