Hull-White calibration for swaptions using neural networks

Alvarez Avamilano, Hugo; Sacchi Córdova, Riccardo; Señaris, Tomás

Hull-White calibration for swaptions using neural networks

License

Citation

Alvarez Avamilano, H, Sacchi Córdova, R, Señaris, T. Hull-White calibration for swaptions using neural networks. 2022. handle: http://hdl.handle.net/10230/56573

This citation was generated automatically.

Abstract

The purpose of this project is to calibrate the Hull-White model using data on at-the-money payer swaptions and Artificial Neural Networks (ANN) for 100 different combinations of expiries (T0) and tenors (Tn – T0). The calibration of the model parameters a and σ is done in the following way: (i) first, we use market swap interest rates to generate 10,000 combinations of (T0, Tn, a, σ); (ii) then, we calculate swaption prices for each of these combinations using a closed-form function derived from the Hull-White (one factor) model in the form of F: (T0, Tn, a, σ) → Price; (iii) we use ANN to learn the inverse mapping relationship (F−1 : (T0, Tn, Price) → (a, σ)); (iv) we use swaptions market prices as arguments for F−1 to find calibrated a and σ for each expiry and tenor. Finally, we use the calibrated parameters to calculate the theoretical Hull-White prices and compare these prices with the prices observed in the market. The results show that the prices generated are very close to the ones observed in the market, suggesting that the ANN is an effective method for calibrating the model parameters accurately.