I want to train holt model on training data, and then make a forecast on test data without estimating parameters again. On several sites I found this solution, but for ARIMA model:
library(forecast)
m_train <- holt(rnorm(500)+runif(500))
m_test <- holt(rnorm(100), model = m_train)
m_train$model
m_test$model
but when I look at the print I can see that smoothing parameters are different for both models.
