我使用了自动 ARIMA 并得到了这样的结果:
Series: JMB
ARIMA(5,1,4)(2,0,2)[96] with drift
Coefficients:
ar1 ar2 ar3 ar4 ar5 ma1 ma2 ma3 ma4
1.3100 0.2710 -1.0215 0.5572 -0.1527 -0.8652 -0.6309 0.7686 -0.2520
s.e. 0.1384 0.1974 0.0752 0.1208 0.0334 0.1389 0.1371 0.0960 0.0797
sar1 sar2 sma1 sma2 drift
0.5959 0.4010 -0.4792 -0.4338 0.0005
s.e. 0.0382 0.0381 0.0388 0.0363 0.0183
sigma^2 estimated as 0.01521: log likelihood=9835.91
AIC=-19636.59 AICc=-19636.56 BIC=-19522.77
> plot(forecast(fit,h=96), xlim=c(120,155) )
Warning message:
In sqrt(z[[2]] * object$sigma2) : NaNs produced and can not use plot (...) funktion.
除了警告之外,残差也太大了。
可能是 Auto Arima 创建了一个错误的模型,我该如何改进这个模型?