Delta-neutral volatility trading using neural networks

In this paper we propose a methodology to forecast the daily changes of the market volatility (ISD) using neural networks. We define the input variables based on specific literature but using lag and horizon of the training set calculated so that we guarantee the significancy of its correlation with the output variable through time. We use this methodology on data from Telebrás PN Stock options from August 1994 through November 1996. Then, based on out-of-sample projections of the model we simulate a trading volatility strategy, creating Delta-hedged portfolios that result in abnormal returns in the market.