I am new to java. I want to add a new *.class file to classifiers. I do
the following things.
1. use javac to get a new class file myclassifier.class
2. add this file to folder weka\classifier
3. use jar to get myweka.jar
After this I find didn't add my new classifier, I think I need to modify
META-INF\manifest.MF But I don't know how, I don't know how to generate
SHA-Digest and MD5-Digest.
Please help me.
Christian Voegeli wrote:
> Dear all,
> Is there any way to obtain the attribute values of a numeric
> attribute? It only seems possible for nominal and string attributes.
No, because a numeric attribute can potentially be any number. The only
way to do this would be to collect all of the values for the attribute
from a set of instances.
> I tried to make the following code:
> The compiler generated the full code but when I tried to run this code, an
> exception occured by the method buildClassifier.
> Could anyone help me?
Its hard to say what is wrong without more information, such as the
classifier being used, how the instance was constructed, the exception
that was thrown etc.
> Sorry about my ignorance.
> Many thanks for your help!
> All the best
When I do classification using NeuralNetwork
through the Explorer's GUI, the network converges
and I get a good result. Then I take the very same
commands the GUI is showing, and run from
MS-DOS command line:
java weka.classifiers.neural.NeuralNetwork -t tr.arff
-T te.arff -L 0.3 -M 0.2 -N 500 -V 0 -S 0 -E 20 -H a
Then the network does not converge and I get the
confusion matrix showed below. The same procedure
works well when I use different classifiers (AdaBoostM1,
etc.) with the GUI and DOS command line version providing
the same result as expected.
I will try to debug the code, but I am wondering if anyone
has some hint about this matter.
=== Confusion Matrix ===
a b c d e f g h i j <-- classified as
0 0 0 0 0 0 0 0 65 0 | a = IY
0 0 0 0 0 0 0 0 50 0 | b = IH
0 0 0 0 0 0 0 0 43 0 | c = EH
0 0 0 0 0 0 0 0 59 0 | d = AE
0 0 0 0 0 0 0 0 48 0 | e = AH
0 0 0 0 0 0 0 0 38 0 | f = AA
0 0 0 0 0 0 0 0 48 0 | g = AO
0 0 0 0 0 0 0 0 56 0 | h = UH
0 0 0 0 0 0 0 0 63 0 | i = UW
0 0 0 0 0 0 0 0 66 0 | j = ER
Aldebaro Klautau - a.klautau(a)ieee.org
UCSD - ECE - Ph.D. candidate
Is there any way to obtain the attribute values of a numeric
attribute? It only seems possible for nominal and string attributes.
I tried to make the following code:
The compiler generated the full code but when I tried to run this code, an
exception occured by the method buildClassifier.
Could anyone help me?
Sorry about my ignorance.
Many thanks for your help!
All the best
I might have found a bug in AttributeTypeFilter class (in WEKA
version 3.2.1, 3.3). It can produce incorrect results and may be
even cause the program to crash.
I'm able to reproduce the bug with:
java -cp $WEKAHOME/weka.jar weka.classifiers.FilteredClassifier
(weather.nominal-modified.arff is a modified version of
weather.nominal.arff that comes with WEKA)
The program exits after printing "null".
The corrected code that is able to rectify the problem is attached.
(the modified part is commented with "LEE, Yoong-Keok")
I would appreciate if anyone is willing to countercheck that this
bug is reproducible on your machine.
Thanks and regards.
I am currently performing experiments on text classification using Weka,
and I am ionterested on using Stacked Generalization applied to an
heuristic classifier I have bilt and some other classification schemas. I
have read in the papers suggested by the Weka book and documentation
(including the paper by Ting and Witten on "SG: when does it work?") that
the level-1 generalizer would optimally be a linear model. Among the Weka
classifiers able to induce linear models, which ones would you suggest as
level-1 learners for Stacking?
Thank you and best regards
Jose Maria Gomez Hidalgo
Departamento de Inteligencia Artificial
Universidad Europea de Madrid - CEES
28670 - Villaviciosa de Odon - MADRID Tfno: +34 91 664 78 00 Ext. 670
e-mail: jmgomez(a)dinar.esi.uem.es WWW: http://www.esi.uem.es/~jmgomez/