Bias in random forest variable importance measures: Illustrations, sources and a solution

BACKGROUND: Variable importance measures for random forests have been receiving increased attention as a means of variable selection in many classification tasks in bioinformatics and related scientific fields, for instance to select a subset of genetic markers relevant for the prediction of a certa... View More