0

I am trying to use HoltWinters(in R) on the below time series data. As can be seen, it has trend + seasonality. I have 25 monthly data points.

enter image description here

But on checking the object returned by HoltWinters() function, below is what I get.

enter image description here

What is causing the alpha, beta, gamma to be at extremes? What can I do to fix this?

Vash
  • 1,767
  • 2
  • 12
  • 19

1 Answers1

0

looks ok to me.

beta=0 means the fitted trend was good for the entire window of data, rather than needing a smooth adjustment from step to step. your graph looks like a good fit to a line after deseasonalization.

alpha=1 says the trend value does not affect the (unobserved) level -- trend is merely an offset to the (observed) y that steadily increases over time, and does not feed back into the system state.

how did you generate this timeseries?

welch
  • 936
  • 8
  • 12
  • Thanks for the response. This is a monthly plot of one of the KPIs(factual) that I am looking to forecast. I know the data points are few but just wanted to take a stab at it. – Vash Mar 14 '19 at 00:57