Holt's method is for a time series that is not seasonal but does display an upward or downward trend.
The beta parameter controls how reactive the model is to the changes in the trend. Be careful with this setting -- trend models can be unpredictable. A beta value that is fine at one point in time can produce an out-of-control model a few months later.
My rule of thumb is to use a value of 0.20 if I think a model has a trend, and never more than 0.40.
If you use a trend model, I highly recommend also using trend dampening profiles to ease off the trend values further out in the model's forecast horizon.
Hope this helps!
Dan
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.