CPI预测的SVM-ARIMA模型构建与数值模拟

SVM-ARIMA Model Construction and Numerical Simulation of CPI Forecasting

居民消费价格指数(CPI)是宏观经济中的前瞻性指标,为经济政策的制定提供数据支撑,发挥指导作用。文章利用CPI的月度数据构建基于小波分解的SVM-ARIMA组合模型,实现了对CPI的精准预测。首先,对2000—2019年的居民消费价格指数序列进行小波分解;然后,对分解后的居民消费价格指数序列分别利用ARIMA模型和SVM模型进行预测;最后,将预测结果进行整合形成对居民消费价格指数的组合预测模型,并选用2020年的实际CPI月度数据与模型预测数据进行有效性验证。结果表明:组合模型的平均绝对百分比误差(MAPE)与均方根误差(RMSE)分别为0.5383%和0.6604%,相较于ARIMA时间序列模型和SVM模型实现了极大的改进。此外,该组合模型的预测分析框架具有较强的适应性和扩展性,可用于其他相同特征类型的时间序列数据的模拟预测。

The consumer price index(CPI) is a prospective indicator in the macro economy,which provides data support and guidance for the formulation of economic policies.This paper uses the monthly data of CPI to construct a SVM-ARIMA combined model based on wavelet decomposition to achieve accurate prediction of CPI.First,the CPI series from 2000 to 2019 is decomposed by wavelet.Then,ARIMA model and SVM model are used to predict the decomposed CPI series.Finally,the above prediction results are integrated to form a combined prediction model for CPI,and the actual monthly CPI data in 2020 and the model forecast data are selected for validity verification.The results show that the MAPE and RMSE of the combined model are0.5383% and 0.6604%,respectively,which achieve a great improvement compared with ARIMA time series model and SVM model.In addition,the predictive analysis framework of the combined model has strong adaptability and scalability and can be used to simulate and predict other time series data with the same characteristic.