Hi to All,
could you please say to me how do the following parameters are
evaluated, could you write for me their formulae?
Mean absolute error
Root mean squared error
Relative absolute error
Root relative squared error
I am trying to open up a data set that has 280000 rows and 23 columns and
create associaon rules using Apriori. I am getting an "out of memory - use
smaller datset or increase heap size" error. I looked at wekalist archives
and found similar issues. The solution given for those who run Weka from the
start menu was to edit the Runweka.bat file.
Following those instructions i edited RunWeka.bat file to change
java -Xmx128m -classpath "%CLASSPATH%;weka.jar" weka.gui.GUIChooser
java -Xmx800m -classpath "%CLASSPATH%;weka.jar" weka.gui.GUIChooser
I am still getting the same error.
Is Runweka.bat the right place/the only place where i need to make this
The reason i am asking is - I tired clicking on Runweka.bat file and i got
a "class not found" error" and the GUI did not open up. So I am wondering
if this is the actual file that runs when i click on Weka from the start
menu or is something else happening too.
I am unfamiliar with running java statements from the command prompt. I
tried copy pasting some commands to increase heap size from the command
prompt and i always got some error in "main" function that i dont
Any help will be greatly appreciated.
I wanted to draw the ROC curve from Weka for my algorithm and dataset.
I did not see a curve but I saw the ROC total area.
I made a 10-fold cross validation and I got the result list.
When I look at the Detailed Accuracy by Class, I see TP-FP values.
I just right clicked the result and chose Visual Threshold Curve->Yes (for
my domain) and
I only see 2 points. One of them is the point I see from Detailed Accuracy
by Class info and other
one is (1,1). And third one is (0,0) but it is not stated explicitly.
And weka calculates the area by using 3 points I explained. But; is it
enough for a problem to use these
3 points to say that an algorithm is better than other one?
How can I change the false positive rate (FPR) and increase it until 1?
Generally in some documents;
ROC curve is shown by using more values not only 3 values.
I think it depends on the threshold value and i did not see any way to
change it. I think the Detailed Accuracy by Class
values are calculated using threshold as 0.5 and it is not allowed to change
it in somewhere.
How can I see more values in that curve or is it feasible to use 3 values
for ROC curve?
P.S: I started to use Weka in these days and I am not professional in Weka.
I read the old messages but could not get any relavent answer.
Sorry if this is answered before or a simple question for group.