Modulus-based successive overrelaxation iteration method for pricing American options with the two-asset Black-Scholes and Heston's models based on finite volume discretization (Q2078260)

In this paper, two fully discrete finite volume methods (FVMs) are studies for pricing the two-asset Black-Scholes and Heston's stochastic volatility American options models. Based on barycenter dual partition, one semi-discrete and two fully discrete backward Euler and Crank-Nicolson FVMs are developed. The linear complementary problems (LCPs) arising from American options pricing are solved with the modulus-based matrix splitting (MSOR) iteration methods, e.g. MSOR method coupled with the FVM. Numerical experiments confirm that MSOR method is more efficient and robust than the classical projected successive overrelaxation (PSOR) method.