2

Before fitting an ARIMA model, we found the trend of the ts is clearly nonlinear, so we remove the trend by subtracting the moving average (calculated by moving a sliding window over the ts and compute the mean of the numbers in the window) from the original data, after which the ts becomes stationary. But, how should I make a prediction in this case?

user112758
  • 758
  • 5
  • 13

1 Answers1

1

Fit the ARMA model as your series is stationary now then you need to add earlier subtracted value ie. moving average to get forecasting of original data.

Arpit Sisodia
  • 1,175