美式期权定价的数值解-概率论与数理统计专业毕业论文.docxVIP

AbstractTitle：Numerical Abstract Title：Numerical Solution ofAmerican Option Valuation Major：Probability and Mathematical Statistics Name：Liang Jiawen Supervisor：OU Jingqi Professor Abstract American option valuation is more complicated than European option valuation， because American options can be exercised at any time up to the expkation date．This involves“exercise boundary”and finite—horizon optimal stopping problems．Because of its complicacy,there are generally no analytical solutions fur American option valuation．Even thou曲we can find this analytical solution，its complicacy of calculation makes it difficult to be applied in practice，so we have to use numerical simulated methods． This paper discusses two numerical methods fur American option valuation．The fjrst js additional model，We use the combination of additiclnal model and a decomposition formula of American option valuation to obtain the approximation of American option value．The second is Least Square Monte-Carlo method．It can obtain the approximation of American option value by approximating to the conditional expected function and following the optimal strategy．This was first proposed by Longstaff and Schwartz in 2001．What this paper is unlike Longstaff and Schwartz’s is，this paper uses second kind Chebyshev polynomials and Legendre polynomials as basis functions．Moreover,this paper will further extent and discuss the convergence of LSM approximations． The main contribution of this paper is：Fkst，propose how to make use of additional model to price American options and embody this method；second， Monte—Carlo methods is proved to be mean square convergent to American option cash flow conditional expected function when it is in the case of two-period exercise dates and converge to this conditional expected function in probabilRy when it is in the case of multi-period exercise dates by using second kind Chebyshev polynomials and Legendre polynomials as basic functions respectively．Finally,this paper compa