Bitcoin Price Prediction Toolkit

This toolkit implements multiple approaches for predicting Bitcoin price movements. It includes traditional statistical methods, machine learning algorithms, and a novel visual CNN approach that uses candlestick chart images.

Implemented Models

1. Visual CNN Approach: Converts candlestick charts to images and uses convolutional neural networks to predict price movements
2. LSTM Approach: Uses Long Short-Term Memory networks to capture temporal patterns in the time series data
3. ARIMA Approach: Applies traditional statistical time series modeling
4. XGBoost Approach: Employs gradient boosting with technical indicators
5. Random Forest Approach: Utilizes ensemble learning with technical indicators and lagged features

Setup and Installation

Prerequisites

• Python 3.7+
• pip (Python package manager)

Installation

1. Clone this repository:

git clone https://github.com/yourusername/bitcoin-prediction-toolkit.git
cd bitcoin-prediction-toolkit

2. Install the required packages:

pip install -r requirements.txt

Usage

Running Individual Models

Each model can be run independently:

# Run the Visual CNN approach
python visual_cnn/visual_cnn_approach.py

# Run the LSTM approach
python lstm/lstm_approach.py

# Run the ARIMA approach
python arima/arima_approach.py

# Run the XGBoost approach
python xgboost/xgboost_approach.py

# Run the Random Forest approach
python random_forest/random_forest_approach.py

Comparing All Models

To run all models and compare their performance:

python compare_models.py

This will:

1. Fetch Bitcoin historical data (if not already downloaded)
2. Run all five prediction approaches
3. Generate comparison charts and tables
4. Save the results to the 'results' directory

Project Structure

bitcoin-prediction-toolkit/
├── btc_prediction_toolkit.py   # Common utility functions
├── compare_models.py           # Script to run and compare all models
├── requirements.txt            # Dependencies
├── README.md                   # This file
├── data/                       # Directory for data files
├── models/                     # Directory for saved models
├── results/                    # Directory for results and visualizations
├── visual_cnn/                 # Visual CNN approach
├── lstm/                       # LSTM approach
├── arima/                      # ARIMA approach
├── xgboost/                    # XGBoost approach
└── random_forest/              # Random Forest approach

Customization

You can customize the models by modifying their parameters:

• Window Size: Number of previous days to consider (affects all models)
• Prediction Horizon: How many days ahead to predict (default is 1 day)
• Technical Indicators: Add or remove indicators in btc_prediction_toolkit.py
• Model Hyperparameters: Tune parameters in each model's script

Results

Results will be saved in the 'results' directory, including:

• Model comparison metrics (accuracy, precision, recall, F1 score)
• Prediction visualizations for each model
• Feature importance charts for applicable models

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Data is fetched from Yahoo Finance using the yfinance library
• Code is organized for educational purposes and can be extended for research or practical use