Table 2. Mean test AUC for different algorithms using BootRank.
Disease/algorithm | T1D | T2D | BD | CD | CAD | RA | HT |
Support vector machine (SVM) | 0.90 | 0.76 | 0.78 | 0.64 | 0.63 | 0.71 | 0.61 |
Random forest (RF) | 0.88 | 0.76 | 0.77 | 0.65 | 0.68 | 0.71 | 0.64 |
Regularized logistic regression (RLR) | 0.91 | 0.77 | 0.76 | 0.696 | 0.71 | 0.78 | 0.68 |
Naïve Bayes (NB) | 0.77 | 0.83 | 0.83 | 0.67 | 0.72 | 0.71 | 0.68 |
Allele count (AC) | 0.80 | 0.79 | 0.80 | 0.63 | 0.59 | 0.65 | 0.61 |
Log Odds (LO) | 0.81 | 0.81 | 0.81 | 0.699 | 0.69 | 0.71 | 0.67 |
Robust adaboost (RAB) | 0.89 | 0.78 | 0.78 | 0.695 | 0.75 | 0.75 | 0.71 |
Majority (all algorithms) | 0.90 | 0.82 | 0.83 | 0.70 | 0.72 | 0.74 | 0.68 |
4-Majority (only RF, RLR, NB and RAB) | 0.91 | 0.82 | 0.82 | 0.71 | 0.75 | 0.77 | 0.70 |
Shown are the average AUC values for test individuals for the different algorithms when using BootRank, or when combining all 7 algorithms (Majority), or only 4 algorithms (4-Majority).