Figure 5.

AUC scores for classifiers tested using different feature combinations. Each bar denotes clinical trial prediction result (AUC score) using one type of feature engineering method (or combination). Each group of bars (cluster) denote performance of one type of classifier. A single red star above a bar indicates a statistical difference with $p<0.05$, two red stars indicate $p<0.01$, compared to the models trained using all features.