How krishgarg344/crypto-price-prediction works
This repository is a notebook-based BTC-USD forecasting project. Its documented workflow turns historical market data into engineered features and 120-day sequences, learns 64-dimensional temporal embeddings with a stacked LSTM, and uses XGBoost to produce predictions and 7-day forecasts. The README also documents randomized hyperparameter search, 60 cross-validated training runs, and reported evaluation metrics. The notebook is the only source file listed, but no notebook excerpts were provided, so the graph reflects README-documented relationships rather than inspected implementation details.
Main entry points: Crypto_Price_Prediction_(LSTM+XGBoost).ipynb XGBoost regressor
Components 9
Data preparation
- BTC-USD history
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - Market features
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - 120-day windows
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb
Hybrid modeling
- LSTM encoder
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - Latent embeddings
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - XGBoost regressor
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - Randomized search
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb
Forecasting and evaluation
- Model evaluation
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb - 7-day forecast
Crypto_Price_Prediction_(LSTM+XGBoost).ipynb
Connections 8
- BTC-USD history → Market features: derive features
README.md - Market features → 120-day windows: form windows
README.md - 120-day windows → LSTM encoder: learn patterns
README.md - LSTM encoder → Latent embeddings: extract embeddings
README.md - Latent embeddings → XGBoost regressor: predict prices
README.md - Randomized search → XGBoost regressor: tune regressor
README.md - XGBoost regressor → Model evaluation: assess predictions
README.md - XGBoost regressor → 7-day forecast: forecast prices
README.md