This is GSoC2012 fork of 'Mothur'. We are trying to implement a number of 'Feature Selection' algorithms for microbial ecology data and incorporate them into mother's main codebase.
With default parameters too many of the low abundance OTUs are being ranked as very important. Changing the stdthreshold to a higher value like .5 gets rid of these OTUs. Suggest implementing a different default parameter or implement mean decrease in accuracy or gini index for feature selection similar to the R implementation.
With default parameters too many of the low abundance OTUs are being ranked as very important. Changing the stdthreshold to a higher value like .5 gets rid of these OTUs. Suggest implementing a different default parameter or implement mean decrease in accuracy or gini index for feature selection similar to the R implementation.