Type of time series to use Consider the following time series plot:

I am trying to fit a time series to this data. I am using it for the purposes of prediction. What type of time series would work here?
I would like to fit a time series to only the part of the data that doesn't spike immediately upward. 
 A: Is the issue something like this: "I want to use a model before the big uptick, and then forecast forward through the uptick -- to see whether the uptick would be predictable from what came before, or was unusual (for some probability level of "unusual")."
If that is the case, I would follow the lead of Makridakis and Hibron (International Journal of Forecasting, 2000, page 459, in the M3 forecasting competition) and run single exponential smoothing, Holt and dampen -- plus their average.  Each of these has a calculation for likely forecast error.  This will tell you how extreme this spike is on the basis of 3 forecasting methods in very common use.
If this works, thank a sailor -- all these methods have a strong historical connection with the U.S.Navy.
A: An exponential smoothing model might work.  I'd say the primary question is: do you expect the massive uptrend to continue or do you expect the series to revert to the mean?  Has anything like this happened in the past or for similar time series?
