无法从 ARIMA 模型中识别 RMSE

Question

我想知道 ARIMA 的误差结果，如 RMSE 等。我有 45 个月的夜灯数据。我有这样的 ARIMA 模型

fitARIMA <- arima(newdata, order=c(0,0,0),seasonal = list(order = c(1,0,0), period = 12))
summary(fitARIMA)
### here is the result
Call:
arima(x = newdata, order = c(0, 0, 0), seasonal = list(order = c(1, 0, 0), period = 12))
            
Coefficients:
       sar1  intercept
      0.4770   572.1038
s.e.  0.1608    38.5140
            
sigma^2 estimated as 26880:  log likelihood = -294.88,  aic = 593.76
            
Training set error measures:
              ME RMSE MAE MPE MAPE
Training set NaN  NaN NaN NaN  NaN
Warning message:
In trainingaccuracy(object, test, d, D) :
test elements must be within sample

谁知道为什么会这样，如何解决这个问题？谢谢

这是我使用的数据

         Jan     Feb     Mar     Apr     May     Jun     Jul
2015  467.38  441.67  579.30  600.41  793.38  576.80  741.21
2016  516.02  241.41  443.20  502.98  497.31  668.08  596.89
2017  325.89  253.30  737.37  462.75  609.31  559.05  581.16
2018  428.74  584.53  508.92  655.63  867.83 1059.98  509.34
         Aug     Sep     Oct     Nov     Dec
2015  634.66  582.00  661.35  249.46  482.33
2016  686.76  598.28  598.23  391.71  492.66
2017  680.36  753.18  476.41    3.12  608.01
2018  820.85  825.13

Answer 1

将 arima 替换为 Arima。 forecast 包中的 Arima() 函数保存由 summary() 函数使用的附加信息。

假设您的 newdata 对象属于 ts class，频率设置为 12，则以下更简单的代码应该可以工作。

fitARIMA <- Arima(newdata, order=c(0,0,0), seasonal = c(1,0,0))
summary(fitARIMA)

无法从 ARIMA 模型中识别 RMSE

Cannot identify RMSE from ARIMA model

statistics

r

time-series

dataframe

arima