[R] Time Series w/ daily or stochastic observation prediction

Community Community at MMHein.at
Wed Nov 2 18:14:15 CET 2011


I've got two different types of behaviors for which I have to predict the future development.

One is a set of historical data with daily observations and which are following some kind of a seasonal pattern, the second one a set of historical data (measure points), with two observation occurring each week, consisting of ten unique (within one single observation) values in parallel, and which doesn't appear to follow any seasonal pattern.

The former one I've aggregated on a monthly level and calculated a prediction model based on Holt Winters, which appears to be pretty nicely fitting when comparing the predicted history with the actual one, and now I have to accomplish the very same on a daily basis (including February 29th in case of a leap year).

The latter one I have not clue at all as how to prepare the data and run it with any potetially qualifying model.


More information about the R-help mailing list