Research On Iron Ore Price Prediction Model Based On Chaotic Time Series And Machine Learning

In the context of deepening reform and opening up,domestic and foreign markets are closely linked.my country’s huge iron ore import transaction volume and its huge transaction volume are an important part of my country’s economic lifeline.Its price risk affects domestic iron ore transactions and iron and steel transactions.The market possibilities are increasing day by day.As a resource-based commodity,iron ore is affected by various aspects and fluctuates more and more frequently than other minerals.Therefore,the price of iron ore is a complex and dynamic system.The work of this paper includes the analysis of the production of iron ore prices at home and abroad,the study of the dynamic system behavior of iron ore prices,and the establishment of a combination of chaotic time series models and machine learning algorithms to predict iron ore prices.Using the time series ARIMA model to extract the linear characteristics of the iron ore price series,the residual time series obtained by fitting the ARIMA model to the linear characteristics is subjected to chaotic analysis,and the phase space reconstruction and chaotic identification of the residual time series are carried out.The reconstructed sequence is input into Long Short Term Memory(LSTM)neural network algorithm and Support Vector Machine(SVM)for prediction,and finally the model results are compared.Finally,the role of iron ore price forecasting in practical applications is explored,and the conclusions are as follows:1.The ARIMA(3,1,3)model was constructed,and the original price series was predicted in a single step.It was found that the effect of fitting and predicting iron ore prices by a single ARIMA time series model was not very satisfactory.Linear features in the sequence,but not sensitive to nonlinear features.2.The chaotic analysis of iron ore prices found that iron ore prices contain chaos,and machine learning algorithms can better capture the nonlinear characteristics of iron ore prices.The forecast effect of ore prices is good.The effect of using LSTM neural network is slightly better than SVM.3.Using chaotic time series and machine learning combined forecasting model to actually forecast iron ore prices at home and abroad,and analyze the medium and long-term supply risks of iron ore.