Hi,           I am trying to use MLearn in MLInterfaces package to do randomforest, clustering, knn etc. How do I predict on a test set for which I do not know the classes? My training set has two classes. Thanks, Tulgan __________________________________________________________________ Yahoo! Canada Toolbar: Search from anywhere on the web, and bookmark your favourite sites. Download it now at http://ca.toolbar.yahoo.com. [[alternative HTML version deleted]]

That's a great question! I don't think MLInterfaces has an easy approach for that at a high level, but one could be added for the next release which is coming soon. Perhaps MLPredict(). In the mean time, you can always get the "native" output of a classifier's result using RObject. (If you are using a genuine xvalSpec you will need to apply RObject at two levels, that should be documented in the vignette.) If that object comes from a family of fitting tools that have a predict(... newdata=...) idiom, then you should be able to go forward without too much effort. A more cunning approach with the software in its current state, for supervised learning only, is to "forge" class labels for the unclassified samples -- just fake them in the pData of the ExpressionSet. Then supply a training set index vector at the xvalSpec parameter that includes only the labeled samples-- the unlabeled samples will not be used to build the classifier, and the testPrediction components of the output object will have bona-fide predictions on these. Of course be sure to ignore the forged labels in any downstream interpretation -- remove them at once! On Fri, Mar 13, 2009 at 12:11 PM, Tul Gan <tulgan@ymail.com> wrote: > Hi, > I am trying to use MLearn in MLInterfaces package to do > randomforest, clustering, knn etc. How do I predict on a test set for which > I do not know the classes? My training set has two classes. > > Thanks, > > Tulgan > > > __________________________________________________________________ > Yahoo! Canada Toolbar: Search from anywhere on the web, and bookmark your > favourite sites. Download it now at > http://ca.toolbar.yahoo.com. > [[alternative HTML version deleted]] > > > _______________________________________________ > Bioconductor mailing list > Bioconductor@stat.math.ethz.ch > https://stat.ethz.ch/mailman/listinfo/bioconductor > Search the archives: > http://news.gmane.org/gmane.science.biology.informatics.conductor > [[alternative HTML version deleted]]