Implied Volatility Surface Estimation Using Transductive Gaussian Fields Regression

Implied volatility estimation is one of the fundamental tasks for asset pricing and risk management. In this paper, we investigate the applicability of semi-supervised regression techniques to estimate an implied volatility surface from the real market option data. Specifically, we employ a transductive Gaussian field regression method since it is able to predict a distribution of the implied volatilities for unlabelled data using only partially labeled data. We've conducted simulation on S&P 500 index data before and after the global financial crisis with discussions of the observed empirical properties of the method.