Ensembles of Dipolar Trees for Prediction of Survival Time
Malgorzata Towska · 2007
In the paper, the application of random forest for prediction of survival time is presented. The observed data loss function is based on inverse probability of censoring weights. The random forest consists of the sequence ofmultivariate regression trees created on the base of the learning sets, randomly generated from the given dataset. The applied regression trees use minimization of dipolar criterion function for finding the splits in the internal nodes.