Fig. 3
From: Rapid diagnosis and comprehensive bacteria profiling of sepsis based on cell-free DNA
The performance of a Random Forest classifier with balanced subsampling for identifying sepsis samples and healthy samples. a The out-of-bag error converges to 0.16, if the number of decision trees is over 100. b The average AUC curves for our diagnosis strategy (red) and a logistic regression scheme (blue) based on the one-third of the samples reserved for testing the model. c The AUC curves of our diagnosis strategy (red) and a logistic regression scheme (blue) based on an independent dataset for validating the proposed algorithm