Basis Prediction And VaR Estimation Of Sse 50 Stock Index Futures Based On Machine Learning

In recent years,economic globalization has suffered setbacks,economic growth of various countries has slowed down significantly,and volatility in global financial markets has intensified.In particular,the sino-US economic and trade frictions that broke out in early 2018 have greatly adversely affected the already weak expectations of capital markets.Therefore,the focus of the majority of investors began to focus on financial risks from the pursuit of income,have turned their attention to futures and other varieties,hoping to achieve the purpose of risk hedging through it.However,as the spot and futures markets are basically independent markets,their trend is largely determined by their respective transaction conditions,so the risk of hedging failure is likely to occur in the process of hedging.Therefore,the hedging strategy can be adjusted in time by studying the basis risk of the hedging of stock index futures,forecasting the basis and predicting the direction and size of the change of basis.Guided by VaR at risk theory,this thesis combines non-parametric kernel estimation method with basis at risk to construct a basis at risk model under the framework of kernel estimation.Empirical analysis is conducted by selecting spot and futures data of SSE 50 index.First,descriptive statistics are made on basis series of SSE 50 Index.Secondly,the empirical mode decomposition method(EMD)was used to decompose the time series of the basis error.Then,the long and short-term memory neural network(LSTM)and support vector machine(SVR)were used to predict the basis error.The accuracy of prediction is compared and the prediction result is obtained.Based on the practical application,this thesis will give some suggestions on the basis risk management of futures hedging,so as to better help investors correctly manage potential financial risks by using futures products and improve their risk warning ability.