使用tslm进行预测

Question

I have created a time series linear model using tslm. It is built using 179 observations of weekly data and I want to forecast the next 26 weeks but keep getting the error:

tslm5<-tslm(tsorders~ trend +
          I(trend^2) + Month.number, data=ppc.order.forecasting[1:179,])

forecast(tslm5,newdata=ppc.order.forecasting[180:205,])

Error in `[[<-.data.frame`(`*tmp*`, length(tmpdata) + 1, value = c(1,  : 
  replacement has 179 rows, data has 26'

How do I use the data in rows 180:205 and tslm5 to forecast the next 26 weeks?

Answer 1

Without more information about your package versions, and the nature of ppc.order.forecasting , it is impossible to know what causes your error.

The following example works using the latest version (8.4) of the forecast package.

library(forecast)
library(ggplot2)

ppc.order.forecasting <- cbind(
  trend = 1:205,
  Month.number = rep(1:12,18)[1:205],
  tsorders = rpois(205, abs((1:205)/50 + 2*sin(2*pi*(1:205)/12)))
) %>% ts(freq=12)
tslm5 <- tslm(tsorders~ trend + I(trend^2) + Month.number, 
              data=subset(ppc.order.forecasting, end=179))

forecast(tslm5, newdata=subset(ppc.order.forecasting, start=180)[,1:2]) %>%
  autoplot()