Model Design And Application Of Copper Price Prediction Under Multi-dimensional Factors And Time Series

Copper metal has excellent performance and is widely used in industries such as power,electronic appliances,transportation energy,construction and real estate.Under the market economy,it is affected by supply and demand,national policies and some uncertainties.Therefore,copper prices tend to be more complicated and less regular.This will cause greater losses to manufacturers,processors,suppliers of production factors and consumers.This situation has brought a relatively unfavorable impact on the copper market.Therefore,predicting the fluctuations of my country's copper price series will help maintain the balance of the copper market and avoid price risks for enterprises..The main contents of the research include:(1)Based on global copper prices,explore the development status of China's copper market in terms of demand,supply,policy and market,and analyze the factors that affect copper price fluctuations.(2)Analysis of factors affecting copper prices.Select copper stocks,copper scrap imports,my country's copper output,consumer price index(CPI),RMB exchange rate(CFETS)and production price index(PPI)as independent variables,and copper price data as dependent variables.BP feed-forward neural network model,Elman feedback neural network model,LSTM model and wavelet neural network model are used to build a prediction model to predict copper prices,and explore the prediction of copper prices in each model Applicability in research.(3)Based on the time series,construct the EWT-MI-LSTM combination model.Use the decomposition model to decompose the volatility complex copper price series,and then use the mutual information method to classify the decomposed sequence reorganization;Then use each prediction model to predict the sequence after reconstruction and classification;Finally,the prediction results of each series are superimposed to obtain the predicted value of the final copper price.(4)The two types of models are used to make a comparative analysis of the copper prices in the next six months,and it is found that when the prediction results of solving the more complex volatility problems are not much different,it is necessary to comprehensively consider the prediction results of the two types of models.When the predictions of the two models differ greatly,you can consider using a model with a smaller error to make predictions,and refer to the prediction characteristics of the neural network model to assist in decision-making.