none
Setting for unkown parameters in the Holt-Winters model. RRS feed

  • Question

  • According to the documentation, they already defined that “Unknown parameters are determined by minimizing the squared prediction error” like below.

    ----------------------------

     Holt-Winters Filtering

    Computes Holt-Winters Filtering of a given time series. Unknown parameters are determined by minimizing the squared prediction error.

    --------------------------------

    However, I’m looking for a workaround way that we can adjust unknown parameters by another values such as “mean absolute deviation” and minimum value of other results.

    Is there any possible idea?

    Tuesday, April 17, 2018 7:53 AM

Answers

  • You can use the accuracy() function in the forecast package to compute the MAE, which is almost the same thing:

     

    m <- HoltWinters(AirPassengers, seasonal = "mult")

    model.pred <- forecast(m, h  = 2)

    accuracy(model.pred)

     

    > accuracy(model.pred)

                       ME     RMSE      MAE       MPE     MAPE      MASE      ACF1

    Training set 1.693717 11.20429 8.393662 0.5718813 3.019525 0.2620538 0.2411207

    • Marked as answer by Kate123456789 Wednesday, April 18, 2018 6:32 AM
    Tuesday, April 17, 2018 7:13 PM