Abstract
In this thesis, we propose Fast Fourier Transform (FFT) method to compute VaR. Once we know the characteristic function of natural logarithm of the underlying asset price, we are able to compute VaR by FFT method. This is very powerful because the characteristic function can be computed easily in a lot of models. Using FFT method to compute VaR can efficiently reduce the computational time without losing the accuracy. In order to check the accuracy, we would like to analyze the bound of error between the true value and the numerical computation.