In the first part of this project, our team introduce several machine learning model including Random Forest, XGBoost, Cat Boost, Light GBM and ANN to generate option price. In doing so, we have outlined our data generation process, the corresponding introductions of the different machine learning models and compare their capability with respect to MAE. The result on training set shows that The ANN model displays a superior predictive power for option pricing when compared to other widely-used tree models. In the next part of the project, our team try to calibrate the Heston model used to price the option price in the previous section , in this part we also use ANN, Light GBM and XGBoost for the pricers of Heston parameters and compare the MAE of different models. While we find that the calibrated heston parameters deviate greatly from the initial value, so machine learning pricers may not good enough for the pricing. For further studies, we will try to find the better model/feature for the calibration and test its performance using the real-world market data.
Jupyter Notebook:
This file contains all the codes which were created and used for this project/paper.
Final Presentation:
This is the presentation my team and I presented in the Practicum Competition organized by the Financial Engineering Department at UIUC.