Purchasing Manager Index(PMI)is an important leading indicator of economic trend,especially manufacturing PMI.PMI is an indicator to measure the national manufacturing cycle.It provides a strong basis for the government,enterprises and institutions to judge the correct economic situation and formulate corresponding countermeasures.Based on the historical data of China's manufacturing PMI index,this paper establishes a mixed model with Shanghai-Shenzhen 300 index and makes analysis and prediction.Firstly,the original sequence is decomposed by EMD algorithm,and several intrinsic mode functions(IMFs)and a residual sequence are obtained.Secondly,LSTM is established to forecast all PMI series of low-frequency manufacturing industry,and AR-MIDAS model is used to predict PMI index of high-frequency manufacturing industry by using high-frequency series(manufacturing IMF1 + CSI 300 IMF1).Finally,the PMI model prediction results of high frequency manufacturing industry and all PMI models of low frequency manufacturing industry are combined to obtain the final AR-MIDAS-LSTM combined prediction model results.The results show that EMD can improve the prediction accuracy of the time series model,and EMD-AR-MIDAS-LSTM has a good prediction effect on the future manufacturing PMI. |