Primary Aim
The project invovles a comprehensive 'bake-off' of several models. The key steps include:
- Testing and comparing baseline models: ARCH, GARCH, EGARCH and GJR-GARCH.
- Selecting the best-perfoming asymmetric GARCH model based on AIC and RMSE.
- Developing a standalone LSTM as a deep learning benchmark.
- Constructing the final hybrid model using the winning GARCH component's residuals to train the LSTM.
Technologies Used
Python
Pandas
NumPy
PyTorch / TensorFlow
Scikit-learn
Matplotlib