How are RandomForest and Adaboost probabilities calculated?

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

How are RandomForest and Adaboost probabilities calculated?

Jaime
Hi all,
I am working with several algorithms in order to classify crops from
information given into each plot/parcel. I am using 4 methods: *(1)*
BayesNaives, *(2)* AdaBoostM1 (with J48), *(3)* RandomForest and *(4)* SMO
(with kernel=PolyKernel).

I want, not only the class for each instance but also its probability for
each class. It can be done via command line adding "-distribution". It adds
the probability matrix to the output so you have  the probability for every
class and for each instance.

However, I have not been able to find how each of the mentioned methods
calculates these probabilites.

If I am right,

-*/Bayes Naives/* has the probability implicit in its calculation.
-*/SMO /*takes the distance to the hyperplane that divides the classes and
fits a logisitic (sigmoid) function, so the given "distribution" are those
values rather than real probabilities.
-In */Random Forest/*, intuitively, but not sure, I understand that, given
100 trees, the distribution of probabilities are obtained from the times a
class is voted from each tree and for each instance.
-In the case of */AdaBoost/*, no idea.

Any correction, explanation (intuitive and/or mathematical) with be more
than welcome.

Thanks you all in advance,
Jaime
Jaime



--
Sent from: https://weka.8497.n7.nabble.com/
_______________________________________________
Wekalist mailing list -- [hidden email]
Send posts to [hidden email]
To unsubscribe send an email to [hidden email]
To subscribe, unsubscribe, etc., visit https://list.waikato.ac.nz/postorius/lists/wekalist.list.waikato.ac.nz
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.html