RFC: LogReport Architecture

[Joost van Baal]

On Sun, Jun 24, 2001 at 02:04:35PM +0200, Egon Willighagen wrote:
> On Sunday 24 June 2001 09:54, Egon Willighagen wrote:
> > On Sunday 24 June 2001 09:41, Christoph Lameter wrote:
> > > On Sun, 24 Jun 2001, Egon Willighagen wrote:
> > > > As long as our limited language can do that i have no problems... but i
> > > > know we cannot make it strong enough to do these kind of things without
> > > > the external "R" program...
> > >
> > > If you have the R program that you have to use then there is certainly no
> > > way around it. Otherwise SQL usually itself it strong enough for the
> > > everyday statistics.
> >
> > Ok, and i am not talking about every day statistics... not just averages
> > etc, but linear regression, clustering, time series analysis etc...
> >
> > At the university i study chemometrics, which is application of these kind
> > of techniques to chemical data...
> 
> Let me make clear that i am not going to apply lire to chemical data...
> I am just going to apply mathematical and statistical techniques that i use 
> in my research too, to data taken from DLF logs...
> 
> For example:
> 
> - prediction of number of request/ number of emails sent...
>   (imagine you could calculate how much time you have left
>    before the upcoming server failure due to overload)
> 
> - clustering of www users, e.g. in time, or domain, or both...
>   (e.g. calculate the best time to do an upgrade... not too hard
>    btw.)
> 
> - clustering of users by some characteristic, e.g. usage, or time
>   (imagine you have three clusters of users:
>    1. accidental user
>    2. regular user
>    3. power user
> 
>    by applying advanced statistics you could determine specifics about
>    these three kinds of users...)

This sounds very good indeed!  The idea is to do this using R?  People
using lire should install R to get such data?  (BTW, just found out R
( http://www.r-project.org/ ) is gpl-ed and runs on almost any platform
one could think of, which makes relying on R not too bad.)

( However, please give me some time now to get a release shipped before
tuesday.... :)

Bye,

Joost


-- 
To UNSUBSCRIBE, email to development-request at logreport.org with a subject of
"unsubscribe". Trouble? Send an email with subject "help" to
development-request at logreport.org