mechres/bitcoin-prediction

Crypto & Stock* price prediction with regression models.

Crypto & Stock Price Prediction with Various Neural Networks

Crypto & Stock Price Prediction and Forecasting Toolkit.

Main.py & GUI.py files let's you select model, start & end date, available models are: Lstm, Catboost, Prophet, XGBoost and LGBM.

Update Log:

→ 16.07.2024
LightGBM Update:
- Load & Save added.
- Docstring added.
- Code improvements.
- Config file added.
- Models & plots now save to the corresponding folder.
→ 12.07.2024
Xgboost update:
- Load & Save added.
- Bayesian Optimization added.
- TimeSeriesSplit for data splitting added.
- Config file added.
- Plot save added.
→ 09.07.2024
Prophet Update
- Save & load works
- Multiprocessing added
- Config file added
- Plots now save to plots directory
04.07.2024 →→→ 08.07.2024
- Updated Lstm, added config file.
- Config files moved to a folder.
- Model save now saves to models folder.
- Old Jupyter notebook's deleted.
- Catboost - Save & Load added

Dependencies:

pip install tensorflow yfinance ta scikit-learn pandas numpy matplotlib catboost prophet lightgbm tkinter pickle PyYAML bayesian-optimization

In case of an update breaks something:

• tensorflow = 2.16.1
• keras = 3.3.3
• prophet = 1.1.5
• yfinance = 0.2.40
• ta = 0.11.0
• scikit-learn = 1.5.0
• catboost = 1.2.5
• lightgbm = 4.4.0

LSTM

Long Short-Term Memory (LSTM). It leverages historical price data from Yahoo Finance and incorporates technical indicators (SMA, EMA, RSI) as well as time-based features to improve the prediction accuracy.

CatBoost

Gradient boosting library. It leverages historical price data from Yahoo Finance and focuses on finding the optimal model parameters through grid search for improved prediction accuracy.

Prophet

Meta's forecasting library designed for time series with seasonality and trend components, leveraging volume as a predictor.

XGBoost

XGBoost is an optimized distributed gradient boosting library designed to be highly efficient, flexible and portable.

LightGBM

LightGBM is a gradient boosting framework that uses tree based learning algorithms

Planned Features:

• Graphical User Interface (GUI)
• Save & Load
• Random Forest Regressor
• Gradient Boosting Regressor

Important Notes:

• This is a side project made in free times, models made with this project shouldn't be used for anything other than experimenting.
• At current level models made with this "tool" is not usable for crpyto market but feel free to modify, use, experiment on it.
• Experimental Project: This is a side project for educational purposes and should not be used as the sole basis for investment decisions.

Old Files(For Reference):

These files provide basic implementations of the models, which you can reference for understanding the underlying algorithms and principles:

• Keras-LSTM.ipynb: Illustrates the fundamentals of LSTM models in Keras.
• CatBoostRegressor.ipynb: Shows a simple CatBoostRegressor example with grid search.
• Prophet.ipynb: Demonstrates the basic usage of Facebook's Prophet library.

Screenshots

Contributing:

Contributions and feedback are welcome! Feel free to open issues or submit pull requests to help improve this project. Let me know if you'd like any other modifications!