A new options pricing method: semi-stochastic kernel regression method with constraints

Abstract

This paper presents a unified semi-stochastic kernel regression method for pricing options under general stochastic volatility model. The method combines semi-stochastic sampling for initial asset values with Monte Carlo simulations to construct a least-squares based kernel function regression solution. This approach can not only approximates option prices, but also determines the Greeks of option. The least square problem is augmented with weighted derivative constraints, which enables flexible adjustment of approximate errors for both option prices and Greeks. Numerical results show the efficiency of the proposed method for the Vanilla option and some exotic options: Asian option, Lookback option, discretely monitored Barrier option and the Basket option with several assets under the stochastic volatility model.