Bitcoin Price Prediction with ARIMA and N-BEATS This project focuses on predicting Bitcoin prices using two distinct approaches: ARIMA and N-BEATS. The dataset is sourced from CoinDesk, and we leverage TensorFlow, statsmodels, and other libraries for time series analysis and modeling.
Time series forecasting is a critical business problem with significant financial impact. Classical statistical approaches often outperform machine learning and deep learning methods in this domain. The M4 competition highlighted this disparity, with most top-ranking methods being ensembles of classical techniques. However, a hybrid approach that combines deep learning with classical models, such as Holt-Winters, has shown promise, leading to further exploration of pure deep learning architectures for improved forecasting accuracy and interpretability.
-
Clone the repository:
git clone https://github.com/yourusername/bitcoin-price-prediction.git cd bitcoin-price-prediction -
Install the required libraries:
pip install -r requirements.txt
The dataset is downloaded from CoinDesk and contains Bitcoin prices from October 2013 to May 2021.
The ARIMA model is used for time series forecasting. We search for the best parameters and fit the model on the training data.
N-BEATS is a neural network architecture for time series forecasting.
We evaluate the models using metrics such as Mean Absolute Error (MAE) and Mean Squared Error (MSE).
The generic DL approach performs exceptionally well on heterogeneous univariate TS forecasting problems using no TS domain knowledge. It is viable to additionally constrain a DL model to force it to decompose its forecast into distinct human interpretable outputs.
- To run the project, clone the repository and install the dependencies as described above.
- Run the Jupyter notebook or Python script provided in the repository.
special thanks to my teammate @Reeptide for his immense contribution .