In my thesis I am trying to develop a context-aware framework for the Google
Android platform (for smarphones) which uses large parts of the Java
library. In order to support reasonings, I am thinking of using WEKA. This
way, users developing context-aware applications using my framework can use
the WEKA machine learning algorithm library to include reasoning in their
As I see it, the only problems are performance and Andoird support (or
Performance: Since smartphones are becoming more and more powerful, I think
they are capable of using WEKA. If not, I can put it on a server where
clients can send machine learning requests.
Android Support: Since Android uses large parts of Java, I think WEKA can be
nicely integrated with Andriod SDK. If this is very difficult, I can also
here use a server.
The problem with using a server is that the all application will be very
dependent of it, which is clearly a bummer, but I can live with it.
Do you think I have ignored other major problems?
Thank you for your help
1)I ａｍ ｗｏｎｄｅｒｉｎｇ ｔｈｅ ｍｅａｎｉｎｇ ｏｆ ｎｕｍｂｅｒｓ ｉｎ ｔｈｅ ｌｅａｖｅ ｎｏｄｅ ｏｆ ａ REPtree．
４:68.83 (41.57/1989.69) [24.01/4049.91]
I only know"4" at the beginning is the series number of node, but what do
the following numbers mean?
2) Can the REPtree predict the interval of number, instead of an exact
Or if I want to predict the interval, I should use other algorithm. I would
greatly appreciate if any of you could give me some advices.
View this message in context: http://old.nabble.com/about-REPtree-tp26964707p26964707.html
Sent from the WEKA mailing list archive at Nabble.com.
Could someone give me any idea how logistic classifier deal with numeric missing values?
Thanks in advance.
Hi, I work with weka in NetBeans 6.8 and i not problen with compile this
pakage weka-src.jar my plataform is Windows.
My problem is in add lines in the GUIChoose i don´t understand how i do it?
In the weka Manual.pdf say in the 18.4.5 MultipleClassHierarchie
My classifier is in classifiers the fold is dummy but i not use the
classpath because in use IDE Netbeans.
java -classpath "weka.jar:dummy.jar" weka.gui.GUIChooser
don't add my classifier.
Tanks for your help.
I'm trying to use weka to predit the profit sales of a company.
I already have an arff filecontaing attributes such as profit, date of
the sale(day, month and year), total sold quantity, etc
I'm trying to use linear regression in order to perform this task.
However i'm not able to understand the results i'm obtaining. Can
anyone help me? Am i supose to get only one predicted value or a series
of values through time?
in the literature i read that it is recommened to use a validation-set to
validate the gerneralisation obtained from training with the training set
and to chose the right model.
after that the accurancy may be discussed with a independent test-set.
i want to do that in weka, but i have the following problem:
after building a model with validation i chose a test set - but there is
a new model built on the training set that was the training and validation
set prior. how can i avoid that?
or is it a valid procedure using the validation set for training after a
my idea was using two "test" sets, one for validation and one for testing.
problem: cross-validation not possible. and when i want so split the
set not taking e.g. the last third of a file it would be a lot of copy &
or time for generating a script to create the validation set.
easiest solution would be using the test set for validation/chosing the
model - but that is no "clean" procedure, is it?!
i also found out that i can save a model. but i miss the feature in weka
to run a test-set on a model without generating a model!
i'm using the gui. is it possible there?! if not, how do i do that via
and another question:
what are the best set sizes for having a total of 2500 cases?
i read in older literature training-test 2/3-1/3 and in some newer one
Witten & Frank book has a section on 0.632 bootstrap and suggest it
would useful for small datasets. Is there an implementation of this
crossvalidation method in Weka? I am new to Weka and have not (yet)
fully figured out how to find various methods.
Many thanks for any help.