While both mentioned haplogroup predictors use a Bayesian-Allele-Frequency approach, this YSEQ predictor uses machine learning with the random forest technique.
Machine learning is in its infancy, so this predictor is unlikely to give you better or more precise results than the Bayesian predictor types, but at least you can consider it as a second opinion
with an independent method.
Note that the YSEQ predictor is based on results of YSEQ customers, but it doesn't reveal the underlying STR profiles and original sample donors (they are not even stored on this web server at all).
A computer based random number generator is the origin for creating random decision trees
which are then just tested with a real life truth set. The teaching process simply selects the best decision trees and uses them for the prediction process.
Please consider this beta version as an experiment which is largely untested and which needs a lot of improvements for reliable haplogroup predictions. We hope that when the number of samples increases, more and more
outlier cases will be covered and considered for the prediction. We hope that this tool will become useful for the genetic genealogy community. The YSEQ Haplogroup Predictor comes with no warranty, explicit or implied.