I have the weekly revenue data for an electronics company the decomposed plot of which is as follows:

enter image description here

I have decided to keep the seasonality and apply a suitable forecasting technique. I tried auto.arima:

> Elec <- read.xlsx("C:/Users/Himanshu.raunak/Revenue/Electronics.xlsx", 1)
> Elec$Date <- as.Date(Elec$Date, format="%Y-%m-%d")
> ElecTimeSeries <- ts(Elec$Revenue, frequency=52)
> ElecArima <- auto.arima(ElecTimeSeries)
> plot(forecast(ElecArima))

I get the following plot:

enter image description here

And the following warning messages:

1: In myarima(x, order = c(p, d, q), seasonal = c(P, D, Q), ... : Unable to check for unit roots

2: In myarima(x, order = c(p, d, q), seasonal = c(P, D, Q), ... : Unable to check for unit roots

3: In myarima(x, order = c(max.p > 0, d, 0), seasonal = c((m > ... : Unable to check for unit roots

and so on.

The ARIMA parameters come out to be as follows:

ARIMA(2,1,2)(0,0,1)[52] with drift


       ar1      ar2      ma1     ma2    sma1     drift

      0.5282  -0.0316  -1.3125  0.3225  0.2283  2497.993

s.e.    NaN      NaN      NaN     NaN    0.0728    NaN

sigma^2 estimated as 3.563e+11:  log likelihood=-2931.26

AIC=5876.51   AICc=5877.1   BIC=5899.6

I realize that the AIC values are quite large.

Could you please point out at what I am doing incorrectly (warning messages and large ARIMA parameters) and provide a better solution. Also I need help understanding the ARIMA plot.

  • 1
    $\begingroup$ I think your frequency parameter is wrong here. I guess you want to add weekly seasonality, so you probably should change it to ElecTimeSeries <- ts(Elec$Revenue, frequency=7). Just a side note, it is hard to help you with this without an actual data $\endgroup$ – David Arenburg Jul 1 '14 at 8:58
  • $\begingroup$ Also, please update to the latest version of the forecast package. The warning message is no longer produced for series with seasonal period greater than 48. $\endgroup$ – Rob Hyndman Jul 1 '14 at 14:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.