[prev in list] [next in list] [prev in thread] [next in thread]
List: wekalist
Subject: Re: [Wekalist] Urgent help with predictions
From: Eibe Frank <eibe () waikato ! ac ! nz>
Date: 2019-03-22 3:55:14
Message-ID: F607F714-5BFD-47ED-96C2-F5AA38967349 () waikato ! ac ! nz
[Download RAW message or body]
It depends on the classification model that has been learned. Most classification \
models in WEKA can produce class probability estimates for a given instance, using \
their distributionForInstance(Instance) method. For example, in J48, REPTree, and \
DecisionStump, when the instance being classified ends up in a particular leaf node \
of the tree structure, the class proportions of the instances in the training set \
that are associated with that leaf node will be used to estimate the class \
probabilities: if there are N instances in the training set associated with the leaf \
node, and C_i of them belong to class i, then the probability of that class is \
estimated as C_i / N.
To find out how other WEKA learning algorithms produce their class probability \
estimates, you will generally have to look at the research papers that introduced the \
corresponding learning algorithms.
Cheers,
Eibe
> On 21/03/2019, at 4:26 PM, Andria Lan <andrialan108@gmail.com> wrote:
>
> Eibe, kindly one thing else, do you know what is the way of computing the \
> "prediction" in that output? Is it related to major probability topic or related to \
> the algorithm itself?
> Thank you.
>
> Andria
>
> On Thu, Mar 21, 2019 at 11:15 AM Eibe Frank <eibe@waikato.ac.nz> wrote:
> The "prediction" is the probability estimate for the class value listed under \
> "predicted". (I did not choose the names.)
> Cheers,
> Eibe
>
> > On 21/03/2019, at 4:13 PM, Andria Lan <andrialan108@gmail.com> wrote:
> >
> > Thank you Eibe for the prompt reply. But I don't know how Wek computes the below \
> > result and output it: inst# actual predicted error prediction
> > 1 1:? 1:Iris-setosa 1
> > 2 1:? 1:Iris-setosa 1
> > 3 1:? 2:Iris-versicolor 0.993
> > 4 1:? 1:Iris-setosa 1
> > 5 1:? 3:Iris-virginica 1
> > 6 1:? 2:Iris-versicolor 1
> > 7 1:? 3:Iris-virginica 0.963
> >
> > What does the values 0.993 mean?
> >
> > On Thu, Mar 21, 2019 at 11:06 AM Eibe Frank <eibe@waikato.ac.nz> wrote:
> > The output is misleading. It does not calculate error indicators in this case. \
> > You will never get the little "+" that is used to indicate errors when labels are \
> > present in the test set.
> > Cheers,
> > Eibe
> >
> > > On 21/03/2019, at 2:00 PM, Andria Lan <andrialan108@gmail.com> wrote:
> > >
> > > Hi all,
> > >
> > > I have loaded both training and test sets into Weka The test set has no \
> > > labels. Therefore, the algorithm labeled them as follows:
> > > === Predictions on test set ===
> > >
> > > inst# actual predicted error prediction
> > > 1 1:? 1:Iris-setosa 1
> > > 2 1:? 1:Iris-setosa 1
> > > 3 1:? 2:Iris-versicolor 0.993
> > > 4 1:? 1:Iris-setosa 1
> > > 5 1:? 3:Iris-virginica 1
> > > 6 1:? 2:Iris-versicolor 1
> > > 7 1:? 3:Iris-virginica 0.963
> > >
> > > 1- My question is that how Weka computed the error prediction for the upper \
> > > results without having the ground truth labels?
> > > 2- Why Weka was able to compute the error prediction compute for the upper \
> > > section, but did not do that in the summary result below?
> > > === Summary ===
> > >
> > > Total Number of Instances 0
> > > Ignored Class Unknown Instances 7
> > >
> > > === Detailed Accuracy By Class ===
> > >
> > > TP Rate FP Rate Precision Recall F-Measure MCC ROC Area PRC Area \
> > > Class ? ? ? ? ? ? ? ? \
> > > Iris-setosa ? ? ? ? ? ? ? \
> > > ? Iris-versicolor ? ? ? ? ? ? \
> > > ? ? Iris-virginica Weighted Avg. ? ? ? \
> > > ? ? ? ? ?
> > > === Confusion Matrix ===
> > >
> > > a b c <-- classified as
> > > 0 0 0 | a = Iris-setosa
> > > 0 0 0 | b = Iris-versicolor
> > > 0 0 0 | c = Iris-virginica
> > >
> > >
> > >
> > >
> > > 3- Does that error prediction here means sort of probability or what?
> > >
> > > Thanks in advance.
> > >
> > > Andria
> > > _______________________________________________
> > > Wekalist mailing list
> > > Send posts to: Wekalist@list.waikato.ac.nz
> > > To subscribe, unsubscribe, etc., visit \
> > > https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> > > http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
> >
> > _______________________________________________
> > Wekalist mailing list
> > Send posts to: Wekalist@list.waikato.ac.nz
> > To subscribe, unsubscribe, etc., visit \
> > https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> > http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html \
> > _______________________________________________ Wekalist mailing list
> > Send posts to: Wekalist@list.waikato.ac.nz
> > To subscribe, unsubscribe, etc., visit \
> > https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> > http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
>
> _______________________________________________
> Wekalist mailing list
> Send posts to: Wekalist@list.waikato.ac.nz
> To subscribe, unsubscribe, etc., visit \
> https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html \
> _______________________________________________ Wekalist mailing list
> Send posts to: Wekalist@list.waikato.ac.nz
> To subscribe, unsubscribe, etc., visit \
> https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
> http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@list.waikato.ac.nz
To subscribe, unsubscribe, etc., visit \
https://list.waikato.ac.nz/mailman/listinfo/wekalist List etiquette: \
http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic