Embedding covariate adjustments in tree-based automated machine learning for biomedical big data analyses

摘要

Boxplots for the results of 100 runs of TPOT on the TG-GATEs data set. Each point corresponds to one run and resides in the boxplot for the pathway (Feature Set) selected in the optimal pipeline for that run. The number above each boxplot indicates the fraction of runs where that pathway was selected. a Results for resAdj TPOT; the y-coordinate indicates the R2 on the hold-out Testing dataset. b Results for classic TPOT; the y-coordinate indicates the balanced accuracy on the hold-out Testing dataset