📈 Bitcoin Price Prediction with LSTM

Forecasting Bitcoin closing prices using Long Short-Term Memory (LSTM) networks — with hyperparameter optimization via Optuna.

Authors: Matheus Braga · Philippe Menge · Luis Guilherme Nunes

📌 About

This project uses LSTM (Long Short-Term Memory) neural networks to predict Bitcoin's closing price based on historical data from 2017 to 2023.
The model is trained on the first 80% of the data and evaluated on the most recent 20%, following a strict time-series split to avoid data leakage.

🏗️ Model Architecture

LSTM layers with configurable hidden units
Sliding window approach for sequence creation
Optimized with Optuna (automated hyperparameter search)
Early stopping to prevent overfitting

⚙️ Hyperparameters Tuned (Optuna)

Window_size
Hidden_layer
Epochs
Batch_size

📊 Evaluation

Predicted vs Actual closing price plot
Training and validation loss curves
MAE / RMSE metrics