'What is the proper way to tune a DSHW model?

I am working on a time series problem with two seasonalities(data: https://drive.google.com/drive/folders/1okMUkBj2W2nF9NkoX4igq2-7QP-cgnSO?usp=sharing) and I came across this method called Double-Seasonal Holt-Winters(DSHW) which works very well for data with two seasonalities.

This is the code so far, the forecast looks way better when compared to SARIMA models but the residual acf and pacf plots still show significant correlation at seasonal periods which means the seasonalities have not been captured efficiently.

seas_data = read.csv("CNERURN.csv")
data = ts(seas_data$CNERURN[0:531], frequency = 96) 
t = msts(data, seasonal.periods = c(12, 12*8))

# dswh

fcast <- dshw(t,
              h = 150,
              alpha = 0.30,
              beta = 0.027,
              gamma = 0.88,
              omega = 0.7,
              #lambda="auto",
              phi = NULL)
autoplot(fcast)

forecast

acf pacf

I was wondering if there is a proper way to tune the parameters alpha, beta, gamma and omega since they're the determining factors. I checked https://www.rdocumentation.org/packages/forecast/versions/8.16/topics/dshw to know what each of these parameters control, but am not able to tune them properly(especially alpha). Any suggestions would be appreciated.

r time-series forecasting

Solution 1:^[1]

The dshw() function will estimate all parameters by minimizing the MSE. So just don't specify them and everything will work fine.

fcast <- dshw(t)

Sources

This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.

Source: Stack Overflow

Solution	Source
Solution 1	Rob Hyndman

'What is the proper way to tune a DSHW model?

Solution 1:[1]

Sources

Related Questions

Solution 1:^[1]