Empirical Study Of Screw Steel Futures Price Prediction In Shanghai Futures Exchange

Steel is an irreplaceable material in the process of national development,and its importance in China's economy is self-evident.For a long time,the steel price has been greatly affected by the changing demand of steel companies and the fluctuation of iron ore price.The steel use enterprises,iron and steel traders and iron and steel producers are all faced with huge price risks.The steel futures market has the function of price discovery and risk aversion driving the majority of steel enterprises participate in the futures market to ensure the cost and profits,and reduce the risk to the lowest.But in the futures market,while hedging is important,a large number of investors have made use of futures prices to discover Function,continue to buy and sell futures in order to profit.Therefore,the statistical analysis of steel futures prices is of great significance.In this paper,the daily closing price data of thread steel futures for 10 years are collected,and their prices are predicted by using the following four models.Firstly,the BP neural network method is used to determine the number of hidden layer nodes to predict the closing price in the training set,and then the closing price of the last 10 days of February 2019 is predicted by using the time series method,and the comparison error and iteration number of the training model on the training set are used to predict the closing price of the final 10 days of February 2019.Finally,the prediction method of non-parametric autoregressive model and germ-parameter model are emphatically discussed,and the selection of kernel function and the determination of window width are introduced in detail.As well as two methods of nuclear estimation,it also makes an accurate prediction of closing price.By analyzing and comparing the four models,it is found that the time series method is better than the time series method for the time series data of the daily closing price.However,BP neural network and non-parametric autoregressive model and germ-parameter model are not much different than the time series model,which showsthat these four methods have some research value in the future price prediction of thread steel.