PMI Combination Forecasting Model Based on Singular Spectrum Analysis
To improve the forecasting accuracy of manufacturing purchasing managers' index (PMI), a model combining singular spectrum analysis (SSA), seasonal difference autoregressive moving average model (SARIMA) and support vector regression (SVR) was proposed, considering cyclicality and non-linearity characteristics of PMI. In this model, PMI was decomposed into major components and noisy components with SSA, and two forecasting models were established for two components separately. Due to the advantages of SARIMA model in handling the linear problem and SVR model in the non-linear problem, the major components were modeled using SARIMA and SVR models, and the noisy component using SVR model, hence, the forecasting results of two components were combined into the final forecasting. The final results showed that the SSA-SARIMA-SVR model had the lowest forecasting evaluation index and the best forecasting effect, indicating its application prospect in PMI forecasting.
万方数据
维普
