A Study Of Option Price Prediction Based On Convolution Neural Network

With the rapid development of financial and capital market,financial derivatives occupy an increasing important position in it.As the fundamental derivative products of financial derivatives,options pricing and price forecasting are always important issues to many scholars.Since February 9,2015,Shanghai 50 ETF index option was listed on the Shanghai stock exchange and has been gradually mature,it marked the rapid development of options market in our country.The black-scholes option pricing model,as the widely used pricing formula,took people's attention since 1973 when it was proposed.When compared with the actual financial market price,it will generate large deviation because of many assumptions.As a data-driven model,neural network has been paid a lot of attention and put into application in the field of option pric ing in recent years.In this thesis,we used convolution neural network,as the representative of the deep learning network to predict price.Based on the black-scholes option pricing theory and the convolution neural network,this thesis established the model to predict the option price.Firstly,the appropriate influencing factors were chosen as input of the forecasting option price model.This thesis studied the structure of neural network.BP neural network and convolution neural network were established.Secondly,Shanghai 50ETF index option was selected as the research object.Through Matlab and Python programming,this thesis established two options price prediction model.Two kinds of price forecasting models were used for the option price forecasting.Thirdly,the prediction accuracy of different models was evaluated by the three error indices of MSE,MAE and MSE.Lastly,combing with the price of option theory that determined by the black-scholes option pricing formula,the model structure and parameters were optimized.Based on the results of evaluating the accuracy of the option price prediction model using three error indices,it is known as that option price prediction accuracy of convolution neural network is higher than BP neural network and black-scholes option pricing model still have significant reference value in option pricing.The results of this thesis improve the prediction accuracy of the option pricing,which has great significance for investors to forecast the option price.