headers
Satish Pillai | 1 May 2003 01:07

viewing decision trees. . .

Hello,

I have a problem: when I run Weka on a PC, I can right click on a 
result in the "Result List" window of the Explorer and a new window 
pops up so I can view decision trees, etc.  Now I'm using a Mac.  Well, 
there is no such thing as right-click on a Mac.  The usual substitute 
is apple-click, option-click, or shift-click.  None of these options 
work.  Does anyone know how to trigger the decision tree window?  Any 
info greatly appreciated.

-thanks,
Satish

-----------------------------------------------
Satish Pillai
Division of Biology (0679)
Stein Clinical Research Bldg. 324
University of California, San Diego
La Jolla, CA 92093

Phone:  (858) 552-8585 x7169
Fax:    (858) 552-7445
Email:  satish <at> biomail.ucsd.edu
-----------------------------------------------
Permalink | Reply | 0 comments |
headers
Castellanos, Malu | 1 May 2003 03:46
Picon
Favicon

training instances in each node of the tree

The string returned from the graph() method of class J48 returns info (like the labels) about each node and
branch followed by the name of the relation and a block of attribute definitions as in the following example:

N0->N1 [label="<= 2"]
N1 [label="YES (3.0/2.0)" shape=box style=filled data =
 <at> relation ClassificationProblem

 <at> attribute bin numeric
 <at> attribute vap numeric
 <at> attribute station numeric
 <at> attribute pcr numeric
 <at> attribute Class {YES, NO}

 <at> data

,
]

however, the section  <at> data comes empty, the training instances that go through the corresponding branch
(in the example above, from node N0 to node N1) are not shown. Does anyone knows how to obtain this
information? (i.e., the instances that arrived to each node of the tree).

thanks!

--malu
Permalink | Reply | 1 comment |
headers
Richard Kirkby | 1 May 2003 04:59
Picon
Picon

Re: training instances in each node of the tree

For J48 you want to use the -L (saveInstanceData) option.

As for your other question of graphing the output, WEKA has a built in
tree visualizer (right-click on the entry of the tree you have created
in the Explorer).

The output is a subset of the Dotty format (modified to include the
instance data), for which you can get tools to produce graphs. Here is a
useful link:

http://seclab.cs.ucdavis.edu/~hoagland/Dot.html

Richard
--

"Castellanos, Malu" wrote:
> 
> The string returned from the graph() method of class J48 returns info (like the labels) about each node and
branch followed by the name of the relation and a block of attribute definitions as in the following example:
> 
> N0->N1 [label="<= 2"]
> N1 [label="YES (3.0/2.0)" shape=box style=filled data =
>  <at> relation ClassificationProblem
> 
>  <at> attribute bin numeric
>  <at> attribute vap numeric
>  <at> attribute station numeric
>  <at> attribute pcr numeric
>  <at> attribute Class {YES, NO}
>
(Continue reading)

Permalink | Reply | 0 comments |
headers
Castellanos, Malu | 1 May 2003 05:18
Picon
Favicon

RE:graph with API ( training instances in each node of the tree)

thanks for your reply.
My problem is that I am not using the Explorer, instead I need to use Weka API to interface with my
application. I use the toString() method on the decision tree object and then print the string, which
outputs the textual form of the tree:

vap <= 2: YES (3.0/2.0)
vap > 2
|   vap <= 3: NO (2.0/1.0)
rdi < 5 ...
...

but what I want is to display the graph of the tree from my Java application. Any suggestion about how to do it?

Thanks

--malu

-----Original Message-----
From: Richard Kirkby [mailto:rkirkby <at> cs.waikato.ac.nz]
Sent: Wednesday, April 30, 2003 8:00 PM
To: Castellanos, Malu
Cc: 'wekalist <at> list.scms.waikato.ac.nz'
Subject: Re: [Wekalist] training instances in each node of the tree

For J48 you want to use the -L (saveInstanceData) option.

As for your other question of graphing the output, WEKA has a built in
tree visualizer (right-click on the entry of the tree you have created
in the Explorer).
(Continue reading)

Permalink | Reply | 1 comment |
headers
Richard Kirkby | 1 May 2003 05:31
Picon
Picon

Re: graph with API ( training instances in each node ofthe tree)

Take a look at the weka.gui.treevisualizer.TreeVisualizer class.

Richard
--

"Castellanos, Malu" wrote:
> 
> thanks for your reply.
> My problem is that I am not using the Explorer, instead I need to use Weka API to interface with my
application. I use the toString() method on the decision tree object and then print the string, which
outputs the textual form of the tree:
> 
> vap <= 2: YES (3.0/2.0)
> vap > 2
> |   vap <= 3: NO (2.0/1.0)
> rdi < 5 ...
> ...
> 
> but what I want is to display the graph of the tree from my Java application. Any suggestion about how to do it?
> 
> Thanks
> 
> --malu
> 
> -----Original Message-----
> From: Richard Kirkby [mailto:rkirkby <at> cs.waikato.ac.nz]
> Sent: Wednesday, April 30, 2003 8:00 PM
> To: Castellanos, Malu
> Cc: 'wekalist <at> list.scms.waikato.ac.nz'
> Subject: Re: [Wekalist] training instances in each node of the tree
(Continue reading)

Permalink | Reply | 0 comments |
headers
Mohammed Waleed Kadous | 1 May 2003 08:35
Picon
Picon
Favicon

Announce: TClass -- time series learner using Weka


Hi everyone,

Hope everyone's Weka projects are running well!

I am pleased to announce the availability of TClass -- a multivariate time
series learner -- that builds on Weka.

TClass is appropriate for problems where each training instance is
described by a multivariate time series -- for example, ECG measurements,
speech recognition, environmental sensors, etc. Each training instance is
labelled with a class. TClass can build classifiers for such problems.

Downloads and more details can be found at:

http://www.cse.unsw.edu.au/~waleed/tclass/

Of course, if you have any queries, please do not hesitate to contact me.

Regards,

Waleed.

---
Dr Waleed Kadous, Senior Research Fellow
Smart Internet Technology CRC, School of Computer Science and Engineering
University of New South Wales, Sydney Australia
Tel: +61 2 9385 6922 Fax: +61 2 9385 4936
E-mail: waleed <at> cse.unsw.edu.au URL: http://www.cse.unsw.edu.au/~waleed/
(Continue reading)

Permalink | Reply | 0 comments |
headers
Saurabh Roy | 1 May 2003 20:37
Picon
Favicon

Least Square and Linear Regression class ?

I am trying to use the Linear and LeastSq. Regressor but it gives errors on
the kind of filters (used in the classification example of Message
classfier). Has anyone used this tool in his/her code. Is it possible to
send the code snippet implementing this tool ?

----- Original Message -----
From: <wekalist-request <at> orc.cs.waikato.ac.nz>
To: <wekalist <at> orc.cs.waikato.ac.nz>
Sent: Wednesday, April 30, 2003 8:25 PM
Subject: Wekalist Digest, Vol 3, Issue 2

> Send Wekalist mailing list submissions to
> wekalist <at> list.scms.waikato.ac.nz
>
> To subscribe or unsubscribe via the World Wide Web, visit
> https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist
> or, via email, send a message with subject or body 'help' to
> wekalist-request <at> list.scms.waikato.ac.nz
>
> You can reach the person managing the list at
> wekalist-owner <at> list.scms.waikato.ac.nz
>
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Wekalist digest..."
>
>
> Today's Topics:
>
>    1. Re: Re: WEKA Evaluation - No predictions on cross
>       validations (Bernhard Pfahringer)
(Continue reading)

Permalink | Reply | 0 comments |
headers
alexsee | 1 May 2003 14:23
Picon

Re: WEKA preds from CV...

I used to hack evaluation.java, too, but recently wrote a WEKA tutorial which includes callClassifier -
this is what you want, plus a few extras. Does not seem to be widely known, though:
http://www.oefai.at/~alexsee/WEKA

Cheers, AlexSee
Permalink | Reply | 6 comments |
headers
Aparna Devaguptapu | 2 May 2003 00:33
Picon
Favicon

Can u pls take me off the weka list?


--- alexsee <at> oefai.at wrote:
> I used to hack evaluation.java, too, but recently wrote a WEKA
> tutorial which includes callClassifier - this is what you want, plus
> a few extras. Does not seem to be widely known, though:
> http://www.oefai.at/~alexsee/WEKA
> 
> Cheers, AlexSee
> 
> 
> _______________________________________________
> Wekalist mailing list
> Wekalist <at> list.scms.waikato.ac.nz
> https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist

__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com
Permalink | Reply | 0 comments |
headers
balne | 2 May 2003 02:18
Picon

cross validation results

Hi,
I am applying Desicion table classifier to a data which I created for my
project. The results are bad when i do 10 fold cross validation, which is
just 1.41 %. I am using GUI version of weka. can I get better results.I
had tried all the options with Desicion Table classifier.Can anyone help
me getting better results.

Any help would be greatly appericated.

RAJ
Permalink | Reply | 0 comments |

Gmane