Re: Source code for an ARIMA model

On Jun 27, 4:33 pm, Wit Jakuczun <w...@xxxxxxxxxxxxx> wrote:
Dnia Wed, 27 Jun 2007 12:22:46 -0700
Jay <k_aj...@xxxxxxxxx> napisa³(a):

The data I am dealing with is the number of rows loaded to
a table once every month. I would like to predict the number of
rows to be processed for an upcoming month, and if the actual rows
processed falls below a confidence level of 95% I need to be notified.

Before you decide you need ARIMA model you should do some analysis. I
understand you have made necessary analysis that directed you
into ARIMA models?

Is there a source code out there written in any language which would
take a set of descrete values, and create a ARIMA model. I need the
source code so that i can implement that in my environment. Any
pointers ?

If you can I would recommend using R for this task. R can be embedded
into all general purpose languages.

Best regards
--
[ Wit Jakuczun <W.Jakuczun [at] wlogsolutions.com> ]
[ WLOG Solutions http://www.wlogsolutions.com]
[ Short offer: data mining, operational research,
time series forecasting, consulting ]

Thanks for the response. I have used some statistical tools like
statgraph which proved the efficacy of a ARIMA model to my data. I
guess thats probably because the ARIMA model boils down to a linear or
exponential smoothing functions depending on the data and takes
different forms of other models. Most of my data sets conformed very
well with a max order of 2 restriction. I was thinking about
implementing the whole box jenkins method including the model order
identification/least squares estimation of parameters and validation.
That would defenetely be an overkill for my requirement and my current
job responsibilities. I do have the ability to execute programs on
Unix and use the output in my ETL tool. So if you can throw some light
on R and how to use it and execute it on Unix, I would really
appreciate that.
Thank you.
Jay

.