Fig. 5
From: Classifying mild traumatic brain injuries with functional network analysis
Comparison of classification performance among subnetworks, individual genes, and predefined functionally gene sets from GO or MSigDB. For each size of feature set, five iterations of five-fold cross validation are used to split the dataset, train, and evaluate classifier. The curves show the median of classification performance, measured by the F-scores, and error bars indicate the standard deviation over five cross-validation experiments