statsmodels forecasting using ARMA model

Rachit picture Rachit · Sep 9, 2012 · Viewed 15.2k times · Source

I want to forecast timeseries data. I read in previous posts that module statsmodels has the required tool for using ARMA method for forecasting which is exactly the one I have been looking for. In spite of that I am having trouble in forecasting the data. Can someone explain the various parameters used in the model and/or provide a sample example?

Answer

Josef picture Josef · Sep 11, 2012

The question is very general, for background information Rob Hyndman's link or any text book for time series analysis will be useful.

Skipper Seabold presented a tutorial at the scipy conference that includes an ARMA example

https://github.com/jseabold/tutorial/blob/master/tsa_arma.py

The various methods and options are described in the documentation for ARMA

http://statsmodels.sourceforge.net/devel/generated/statsmodels.tsa.arima_model.ARMA.html

There is currently no book or article style description of the ARMA model in statsmodels that I'm aware of, but maybe you will get more specific answers if you ask a more specific question with what you want to do and what your troubles are.