基于ARMA与LSTM的CPI经济指标分析与预测方法研究

Research on CPI Economic Index Analysis and Prediction Approach Based on ARMA and LSTM Algorithm

CPI是度量消费品和服务项目价格水平变动情况的经济指标,分析预测CPI指数变动对于把握宏观经济运行及政府制定经济政策具有重要现实意义。选取我国相应时间区间的CPI指数数据集CPIData进行时序序列平稳性检验,并采用ACF、PACF和D-W检验进行自相关性检验,验证了CPIData不具有自相关性;采用LSTM算法进一步对CPIData序列进行预测,预测结果与实际数据变化趋势相一致,训练损失小、预测结果更准确。

CPI is an economic index to measure the change of products and service price level.It is of great practical significance to analyze and predict the change of CPI index for grasping the macro-economic operation and making economic policies by the government.The domestic CPI index dataset in specific period is selected and the CPIData stationarity test of time series is carried out.ACF,PACF and D W tests are used to inspect the autocorrelation.It is verified that the CPIData has no autocorrelation in this experiment.Adopting the LSTM algorithm to further predict the CPIData series,the paper finds that the prediction results are consistent with change trend of the actual data,with less training loss and more accurate prediction results.