Since China’s first option,the SSE 50ETF option,was issued on the Shanghai Stock Exchange on February 9,2015,China’s options market has developed vigorously.More and more investors have focused their attention on the options market and began to take advantage of it.Options try a variety of hedging portfolios for asset allocation and risk management,which also makes the option market’s position in the financial market more and more important,whether it is due to government supervision of the financial industry or investors’ influence on market rules and trading products.A rational understanding is particularly important for the study of option market theory.The implied volatility not only reflects the market supply and demand relationship and the degree of transaction risk,but also contains the expected level of market participants for the future volatility of asset prices.It is an indicator that contains traders’ sentiment towards the future market.Therefore,the accurate measurement and prediction of the implied volatility surface is not only conducive to the reasonable pricing of financial derivatives,but also a major difficulty in market transactions and financial risk supervision.Today,when SSE50 ETF options are becoming more and more active,this article chooses to study the implied volatility of SSE 50 ETF options.With the rapid development of artificial intelligence and machine learning technology,deep learning models have performed well in image recognition,medical diagnosis,engineering fault prediction,automatic driving,etc.,and their excellent performance on nonlinear regression problems has been obtained.Verify multiple times.In the field of implicit volatility surface prediction,some deep learning models such as BP neural network and LSTM network have been proved to have good theoretical and practical value.This article uses a deep learning model that is not widely used in the field of implied volatility surface prediction—convolutional neural network to predict the implied volatility of SSE50 ETF options.In this model,we first organize the ten time series factors related to the implied volatility into a panel data set,and use the two-dimensional data matrix as a tensor to train the convolutional neural network to realize the implied volatility After obtaining the forecast data,we use cubic spline interpolation to fit the entire implied volatility surface to get the forecast of the implied volatility surface.The conclusions of this paper are as follows: First of all,the performance of convolutional neural networks is better than BP neural networks and LSTM neural networks in the prediction of implied volatility;in addition,although the use of convolutional neural networks to predict numerical errors is still above 10%(The real data is 12.12%),but the model is very accurate in fitting the trend of implied volatility.The implied volatility predicted by the convolutional neural network is used to predict the overall trend of the surface and the implied volatility smile structure,The term structure is also almost the same as the real market performance;finally,considering the limited amount of data used in this article,the training network is also affected by equipment restrictions,so it is not possible to choose a deeper network to study the implied volatility surface Predict the problem,but the data conclusions in this article can still show that the convolutional neural network has high research value,and the method itself has strong scalability.When the amount of data is sufficient and the computing power of the device is sufficient,the convolutional neural network The ability to extract features of volatility data,the ability to extract time-varying dynamics of implied volatility,and even capture the periodicity of the option trading market can achieve better performance.The innovation of this article is that the current deep learning in feature extraction and index prediction,as well as the processing of high-frequency data and the non-linearity of influencing factors,is being valued by various research fields,but it is currently applied to the depth of implicit volatility prediction The learning network is basically a BP neural network and an LSTM network.Few studies have applied convolutional neural networks to this problem.Therefore,this article explores this field by constructing an effective implicit volatility prediction model based on deep convolutional neural networks,and applies the convolutional neural network method to this field,hoping to predict the implicit volatility surface Make a contribution to the research of the problem. |