Bitcoin Price Forecasting with SARIMA

This project applies the SARIMA (Seasonal ARIMA) model to forecast Bitcoin prices based on historical data. The model performs data preprocessing, stationarity checks, Box-Cox transformation, ARIMA parameter selection, and provides forecasts with confidence intervals. Visualizations of historical data and forecasted prices are also included.

Table of Contents

• Installation
• Usage
• Project Structure
• Dependencies
• Model Explanation
• Visualizations
• License

Installation

To get started with the project, clone this repository to your local machine and install the necessary dependencies.

https://github.com/MohamadPirniakan/Bitcoin-_Price-_Forecasting-_with_SARIMA.git
cd bitcoin-price-forecasting-sarima
pip install -r requirements.txt

Usage

Once you've installed the dependencies, you can run the forecasting script.

python forecast_bitcoin.py

This will:

• Load and preprocess the Bitcoin data.
• Apply the SARIMA model.
• Output the forecast and visualizations.

Make sure to update the dataset path in the script if necessary.

Project Structure

bitcoin-price-forecasting-sarima/
│
├── data/
│   └── crypto-markets.csv  # Dataset with historical cryptocurrency data
│
├── forecast_bitcoin.py     # Main script to run the forecasting model
├── requirements.txt        # List of dependencies
└── README.md               # Project description and usage

Dependencies

This project uses the following Python libraries:

• pandas - for data manipulation and analysis
• numpy - for numerical computations
• matplotlib - for data visualization
• seaborn - for statistical data visualization
• statsmodels - for time series analysis (SARIMA model)
• scipy - for statistical functions
• itertools - for iterating over parameter combinations
• datetime - for handling date and time operations

To install the dependencies, run:

pip install -r requirements.txt

Model Explanation

SARIMA (Seasonal ARIMA) is an extension of the ARIMA model that supports univariate time series data with a seasonal component. In this project, the SARIMA model is applied to Bitcoin price data, with the following steps:

1. Data Preprocessing: The data is cleaned and resampled to a monthly frequency.
2. Stationarity Check: The Dickey-Fuller test is applied to ensure the series is stationary.
3. Transformation: A Box-Cox transformation is used to stabilize variance.
4. Differencing: The series is differenced to achieve stationarity.
5. Model Fitting: The SARIMA model is fitted using various parameter combinations.
6. Forecasting: Future Bitcoin prices are predicted for the next 12 months.

Visualizations

The project generates the following visualizations:

• ACF and PACF Plots: To determine the optimal parameters for the SARIMA model.
• Forecast Plot: Historical Bitcoin prices and the forecasted prices for the next 12 months.
• Confidence Interval: A shaded region representing the uncertainty in the forecast.

License

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