Fig. 6From: Comprehensive study of semi-supervised learning for DNA methylation-based supervised classification of central nervous system tumorsPrediction performance of random forest (RF) and neural net (NN) classifiers at family level when trained with different combination of reference samples and semi-supervised (SS) predicted labeled samples. A Balanced accuracy of RF and NN. B Proportion (left panel) and count (right panel) of high (≥ 10 samples, red) and low (< 10 samples, green) frequency referent labels, high confident (HC) SS labels with high frequency (blue) and low frequency (orange) families, and low confident (LC) SS labels in high frequency (yellow) and low frequency families (purple). Asterisks indicate statistically significant difference performed by Tukey Honest Significant Difference test at 0.05 alpha levelBack to article page