[R] Classification Analysis

ozric@web.de ozric at web.de
Thu Apr 18 08:32:22 CEST 2002


Hi Bill,
a really nice fundus especially for "fuzzy-papers" 
what is in present my main interesting) you can download at http://www.scch.at/index.html . If you have a special method  interest let me now ,beacuse i have got a growing pdf-archive.
....my learning by doing project for R is a fuzzy association rule approach.

suggestion for Rishabh:
>> >All this work I am doing is part of a Phd and my deparment does not really
>> >neural network for situations like these because they say it acts like a
>> >black box and we don't really know what is happening in the inside. So I
>> >would like to concentrate on using pure statistical techniques. 
I'm not complete agree that neural networks are a black box, because
it is possible to measure the sensitivity.

One sophisticated strategy i.e. is 
1. learn a neural network with independent and classification variable
2. If the neural network have got a good learning result test the generalization. 
3.If this is ok for your purporse, change the values for all cases  of the independent variable (which sens. you interested).
4. Now generalize again the neural network with the modified data and you get a new distribution of the classification variable and now the influence of the modification.

The advantage in contradiction to "logistic regression" is that you are not bound to linearity and can i.e. test interaction effects when you change 2 variables at the same time.
The difficulty with ANN here is the estimation of the neural network paramteres for good training and test approximation, but for this exist genetic algorithm.


regards,christian

>Hello Ozric,
>
>I am interested in knowing what papers you have available.
>
>Thanks,
>
>Bill
>
>ozric at web.de wrote:
>
>> Rishabh,
>> yes your "PseudoCode" idea is what i want in R,too.
>> But even i mentioned , i didn't know of something like that in R/S-plus and i have got open eyes for this.
>>
>> Perhaps anybody in the whole R community work on machineLearning-algorithm
>> with rule generating,too ???
>>
>> So you must program a function for yourself (i.e. C4.5 or FS-ID3) !?
>>
>> (If you are interesting for papers (pdf) with algorithm let me know).
>> regards,Christian
>>
>> Am 17.04.2002 15:13:47, schrieb "Rishabh Gupta" <rg117 at ohm.york.ac.uk>:
>> >Thanks for your reply.
>> >I am still learning these aspects of statistical analysis, so if I don't
>> >make sense please forgive me.
>> >All this work I am doing is part of a Phd and my deparment does not really
>> >neural network for situations like these because they say it acts like a
>> >black box and we don't really know what is happening in the inside. So I
>> >would like to concentrate on using pure statistical techniques. Ideally what
>> >I would like is some kind of a function that calculates the "classification
>> >rule" when given a grouped data set :
>> >    rule <- classification( GroupVar ~ DepVar1 + DepVar2 + DepVar3 + DepVar4
>> >..... + ...... DepVarX )
>> >
>> >Then I would be able to use that rule and apply to a single data element for
>> >which the group is not known:
>> >
>> >    theGroup <- rule( DataElement )
>> >
>> >I have looked at the rpart package in R but I am not entirely sure how to
>> >use in a way that I can create the "classification rule" and then use that
>> >rule. I understand that it is a general problem but it's made more difficult
>> >for me because I have to deal with 750 dependent variables.
>> >
>> >Your help is greatly appreciated.
>> >
>> >Many Thanks
>> >
>> >Rishabh
>> >
>> >----- Original Message -----
>> >From: "Huntsinger, Reid" <reid_huntsinger at merck.com>
>> >To: "'Rishabh Gupta'" <rg117 at ohm.york.ac.uk>; <r-help at stat.math.ethz.ch>
>> >Sent: Tuesday, April 16, 2002 5:00 PM
>> >Subject: RE: [R] Classification Analysis
>> >
>> >
>> >> This is a very general problem and a very large area of
>> >statistics/computer
>> >> science/etc is concerned with it. R provides lots of possibilities; you
>> >> might find tree-based approaches (recursive partitioning) to suit your
>> >> needs; in that case, rpart and the new random forests package will be of
>> >> interest. Also see package e1071 and the VR packages for starters. There
>> >are
>> >> lot of other possibilities; you might want to have a look at Ripley,
>> >Pattern
>> >> Recognition and Neural Networks, for example, to see some.
>> >>
>> >> Reid Huntsinger
>> >>
>> >> -----Original Message-----
>> >> From: Rishabh Gupta [mailto:rg117 at ohm.york.ac.uk]
>> >> Sent: Tuesday, April 16, 2002 11:14 AM
>> >> To: r-help at stat.math.ethz.ch
>> >> Subject: [R] Classification Analysis
>> >>
>> >>
>> >> Hi everyone,
>> >>     Could somebody explain to me what is the package/function for
>> >> classification analysis. I am performing analysis of music files in the
>> >form
>> >> of MIDI files. I end up with about 750 dependent variables from the
>> >> analysis, I also have a number of independent/grouping variables that I
>> >set
>> >> manually. What I would like is to be able to predict which group a
>> >> particular MIDI files belongs to given the 750 dependent variables. In
>> >order
>> >> to this I have to perform classification analysis on a sample set of MIDI
>> >> files where I know what group they belong to. I want to extract the
>> >> 'classification rule' that would enable me to predict the group of each
>> >MIDI
>> >> file (there would be a different classification rule for each grouping
>> >> variable). Can anybody explain what is the best way of doing this in R.
>> >What
>> >> is the best package/function that would enable me to perform
>> >classification
>> >> analysis.
>> >>
>> >> Any help would be greatly appreciated.
>> >>
>> >> Many Thanks For Your Help!
>> >>
>> >> Rishabh
>> >>
>> >> -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.
>> >-.
>> >> -.-
>> >> r-help mailing list -- Read
>> >http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
>> >> Send "info", "help", or "[un]subscribe"
>> >> (in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
>> >>
>> >_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._.
>> >> _._
>> >>
>> >>
>> >> --------------------------------------------------------------------------
>> >----
>> >> Notice:  This e-mail message, together with any attachments, contains
>> >information of Merck & Co., Inc. (Whitehouse Station, New Jersey, USA) that
>> >may be confidential, proprietary copyrighted and/or legally privileged, and
>> >is intended solely for the use of the individual or entity named in this
>> >message.  If you are not the intended recipient, and have received this
>> >message in error, please immediately return this by e-mail and then delete
>> >it.
>> >>
>> >>
>> >============================================================================
>> >==
>> >>
>> >> -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.
>> >-.-.-
>> >> r-help mailing list -- Read
>> >http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
>> >> Send "info", "help", or "[un]subscribe"
>> >> (in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
>> >>
>> >_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._.
>> >_._
>> >>
>> >
>> >-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
>> >r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
>> >Send "info", "help", or "[un]subscribe"
>> >(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
>> >_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>> >
>> ________________________________________________________________
>> Keine verlorenen Lotto-Quittungen, keine vergessenen Gewinne mehr!
>>
>> -.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
>> r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
>> Send "info", "help", or "[un]subscribe"
>> (in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
>> _._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._
>
>
______________________________________________________________________________
Sie wollen mehr? Mehr Speicher, mehr Mail, mehr Erlebnis, mehr Leistung, 


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._



More information about the R-help mailing list