Sequential estimation of state-space model parameters
At LUT University, Finland, a recent MSc thesis studies FX trading and the 2015 Swiss Franc (CHF) de-pegging event with SMC^2 (Sequential Monte Carlo Squared) and stochastic differential equations. The event offered possibilities for triangular arbitrage with the triplet (EUR, USD, CHF). Analysis shows that biggest risk-free profits could have been made at the start of the event before rates started to slowly revert towards the equilibrium.
The focus in this thesis is in sequential state and parameter estimation with a generic algorithm SMC^2, which is tailored for non-linear and non-Gaussian models. SMC^2 is a combination of two particle filters. Numerical experiments show that the approach is an effective tool in sequential inference of state-space models. Moreover, the thesis includes common state-space model state and parameter estimation with one physical system and models with stochastic differential equations such as Heston stochastic volatility model.
Publication available at http://urn.fi/URN:NBN:fi-fe2021042111262
Contact for enquiries: prof Lassi Roininen email@example.com