|
Getting class probabilities w/ SMO
|
View:
New views
5 Messages
—
Rating Filter:
Alert me
|
 |
Getting class probabilities w/ SMO
by prog_lady
::
Rate this Message:
First of all, sorry for the naive question i'm new to weka and
to classification so... I have trained text data using SMO to classify as positive and negative when I test an instance, I would like to get the probability of the prediction, I was reading that I could do this using -M however when I use this option for training, when I do the test I still get the output below, which doesn't give a probability of the class prediction, just the prediction itself. What am I doing wrong here??? btw this is the command I am using to build the model... java -Xmx2048m -classpath /Users/lisa/Desktop/weka-3-5-8/weka.jar weka.classifiers.functions.SMO -d /tmp/movie_model.model -c first -t movie_training.arff -M ... + 0.0019 * (normalized) young + -0.0679 * (normalized) your - 0.7147 Number of kernel evaluations: 372592 (97.499% cached) Time taken to build model: 6.07 seconds === Predictions on test set === inst#, actual, predicted, error, probability distribution 1 1:neg 1:neg *1 0 === Evaluation on test set === === Summary === Correctly Classified Instances 1 100 % Incorrectly Classified Instances 0 0 % Kappa statistic 1 Mean absolute error 0 Root mean squared error 0 Relative absolute error 0 % Root relative squared error 0 % Total Number of Instances 1 === Detailed Accuracy By Class === TP Rate FP Rate Precision Recall F-Measure Class 1 0 1 1 1 neg 0 0 0 0 0 pos === Confusion Matrix === a b <-- classified as 1 0 | a = neg 0 0 | b = pos |
|
|
Re: Getting class probabilities w/ SMO
by Mark Hall-9
::
Rate this Message:
On 2/7/09 5:11 AM, prog_lady wrote:
> First of all, sorry for the naive question i'm new to weka and > to classification so... > > I have trained text data using SMO to classify as positive and negative > when I test an instance, I would like to get the probability of the > prediction, > I was reading that I could do this using -M however when I use this > option for training, when I do the test I still get the output below, > which doesn't give a probability of the class prediction, just the > prediction > itself. What am I doing wrong here??? > > btw this is the command I am using to build the model... > java -Xmx2048m -classpath /Users/lisa/Desktop/weka-3-5-8/weka.jar > weka.classifiers.functions.SMO -d /tmp/movie_model.model -c first -t > movie_training.arff -M I can't see a problem with how you are applying SMO. Can you post the command you use to load the model and apply it to your test data? Cheers, Mark. -- Mark Hall Senior Developer/Consultant, Pentaho Open Source Business Intelligence Citadel International, Suite 340, 5950 Hazeltine National Dr., Orlando, FL 32822, USA +64 7 847-3537 office, +64 21 399-132 mobile, +1 815 550-8637 fax, Skype: mark.andrew.hall, Yahoo: mark_andrew_hall Download the latest release today <http://www.sourceforge.net/projects/pentaho> _______________________________________________ Wekalist mailing list Send posts to: Wekalist@... List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html |
|
|
Re: Getting class probabilities w/ SMO
by prog_lady
::
Rate this Message:
Hi Mark,
The command I'm using to test the data is... java weka.classifiers.functions.SMO -l model.model T data.arff -c first -p 0 when I try to use the -M option here I get that it is an illegal option
|
|
|
Re: Getting class probabilities w/ SMO
by Mark Hall-9
::
Rate this Message:
On 2/7/09 10:36 AM, prog_lady wrote:
> Hi Mark, > The command I'm using to test the data is... > java weka.classifiers.functions.SMO -l model.model T data.arff -c first -p > 0 You don't need to add the -M option again. You will need to add the -distribution option (in conjunction with -p) to get the probability distributions output. Cheers, Mark. -- Mark Hall Senior Developer/Consultant, Pentaho Open Source Business Intelligence Citadel International, Suite 340, 5950 Hazeltine National Dr., Orlando, FL 32822, USA +64 7 847-3537 office, +64 21 399-132 mobile, +1 815 550-8637 fax, Skype: mark.andrew.hall, Yahoo: mark_andrew_hall Download the latest release today <http://www.sourceforge.net/projects/pentaho> _______________________________________________ Wekalist mailing list Send posts to: Wekalist@... List info and subscription status: https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html |
|
|
Re: Getting class probabilities w/ SMO
by prog_lady
::
Rate this Message:
I used this command
java weka.classifiers.functions.SMO -l model.model -T data.arff -c first -p 0 -distribution and got the output below, i thought the prediction and distribution would be somewhere between 0 and 1 not 1 itself? I am attaching my training and test files if someone wouldn't mind looking at them :) I am wanting to get the probability of a test instance being positive or negative so that I can classify a sentence as being not only positive or negative but the probability of being positive or negative... movie_review_630.arff testFile8001.arff .... Logistic Regression with ridge parameter of 1.0E-8 Coefficients... Class Variable neg ===================== pred -95.7603 Intercept -101.5845 Odds Ratios... Class Variable neg ===================== pred 0 === Error on test data === Correctly Classified Instances 0 0 % Incorrectly Classified Instances 1 100 % Kappa statistic 0 Mean absolute error 1 Root mean squared error 1 Total Number of Instances 1 === Confusion Matrix === a b <-- classified as 0 1 | a = neg 0 0 | b = pos === Predictions on test data === inst# actual predicted error distribution 1 1:neg 2:pos + 0,*1
|
| Free embeddable forum powered by Nabble | Forum Help |
