Analysis And Prediction Of Option Implied Volatility Based On Machine Learning Algorithm

The implied volatility of options has strong reference value and economic significance for analyzing the potential risks of option market and the volatility of option market in the future.Volatility index reflects investors' expectation of the future volatility of the option market,and can intuitively reflect the implied volatility of the option market.On November 28,2011,Shanghai Stock Exchange released the first China volatility index-China wave index.However,on February 22,2018,China's wave index stopped publishing.In this context,the analysis and prediction of option implied volatility has strong theoretical and practical significance.This paper proposes to use the machine learning algorithm of artificial neural network,random forest and support vector machine to classify and regression predict the implied volatility of Shanghai 50ETF call option.It creatively combines the data-driven machine learning technology with financial prior knowledge,compared with the traditional prediction methods,such as GRACH model,It greatly improves the prediction speed and accuracy of option implied volatility.Firstly,this paper uses the grid search algorithm to determine the model parameters of artificial neural network,random forest and support vector machine,so as to make all kinds of models have the best generalization performance.For regression prediction,the implied volatility obtained by Newton interpolation method is selected as the label value,and the five indexes of execution price,option closing price,underlying asset closing price,risk-free interest rate and remaining maturity time are selected as the feature value.For classified prediction,options with volatility greater than 2%are defined as dangerous options,and options with volatility less than 2%are defined as safe options.Then,three algorithms are used to classify and predict the call options of SSE 50ETF.Finally,according to different prediction types,different evaluation indexes are selected to evaluate the prediction results.The results show that among the three algorithms,random forest algorithm has the best prediction performance whether regression prediction or classification prediction.This paper enriches the relevant work in the field of option random Volatility Prediction,provides a certain reference value for subsequent research,and also provides a new method for Volatility Prediction and option risk early warning.