Modeling And Forecasting Of Implied Volatility Surfaces

As one of the earliest stock index option products in China,SSE 50ETF options have provided many investors with functions of risk management and asset allocation.At the end of 2019,the Shanghai and Shenzhen 300 ETF options were listed,and investors’ demand for stock index options was increasing.Understanding and playing the role of stock index options in depth became a consensus among more and more people.Therefore,this article starts with the SSE 50ETF options and studies the"substitutes" of option prices—implied volatility.By analyzing the current research situation at home and abroad,and by comparing and analyzing the three main methods of implicit volatility surface modeling,it is found that the semi-parametric method has the advantage of being more flexible.On the basis of determining the method,construct the model and filter,select the model with the best performance and estimate the parameters of the cross section,thereby obtaining the implied volatility surface of each cross section in the sample period,and determine these Parameters of surface morphology.After the construction of the implied volatility surface is completed,this paper innovatively proposes to use the LSTM model in data science in the second step of the semiparametric method.As a model in the RNN family,the LSTM model has long-term memory capability for time series,and can theoretically effectively predict hidden volatility surfaces.In the following empirical part,this paper divides the sample into a training set and a test set,trains the LSTM model on the training set,and then uses the daily data to predict the next day’s volatility surface.The results show that the LSTM model can indeed effectively predict the implicit volatility surface of 50ETF options.Further,in this test,the VAR model and ARIMA model that have been used by many scholars are used as a comparison model in the test process.The results show that although the LSTM model is slightly inferior to the former in the performance of the prediction results,it is not much different and is used as Members in learning,LSTM model has great potential for future scalability and optimization.In short,this paper believes that it is feasible to use the LSTM model to predict the implied volatility surface.Although in terms of prediction results,not all discrete points are accurately located on the surface,but when this method is used to predict the direction of change of the volatility surface,the results show that it has certain persuasion.