Research On Options Prediction Based On Neural Network

As a mature and fundamental financial derivative,options can provide risk management,cost control,hedging and other capabilities in the financial market.Option is also a relatively complex financial tool,and the research methods of option pricing are also developing rapidly and becoming more abundant.The two basic components of option value are:embedded value and time value.The assumptions established by the traditional option pricing model may be different from the actual market,resulting in pricing and arbitrage errors in the actual use of the model.Therefore,it is necessary to explore option price prediction models with higher prediction accuracy.Building a high-accuracy option pricing model requires not only understanding and exploring more abundant options price influencing factors,but also finding appropriate research methods.In the financial field,deep learning has been widely concerned,and option pricing is also a research hotspot.The concept of deep learning originates from the study of artificial neural networks.The structure of deep learning models consists of multilayer perceptrons with multiple hidden layers.Neural network is the basis of deep learning.This paper adopts multiple network models in the field of neural network as a method for option price prediction.In this paper,the SSE 50ETF option is the research object,and the daily market volume and price data of the option since the listing of the SSE 50ETF option and the daily market data of the SSE 50ETF are used.Predict the closing price or the highest price of the option based on the opening price of the option,the exercise price of the option,the option validity period and the opening price of the fund.And calculate MSE,RMSE,MAE,MAPE four regression evaluation indicators as model prediction accuracy evaluation criteria.Explore whether the neural network is suitable for option prediction through data experiments,and which neural network algorithm model is better.The results of this paper show that the neural network algorithm model can be used for option price prediction,and the prediction error is small.In the comparison of prediction results of multiple neural network models,the one-dimensional convolutional neural network model has the best experimental effect and the smallest error.