Table 2.
Machine learning model performances for UKB. The best prediction results, based on the optimal hyperparameter values and the optimal number of PCs, are shown for each model.
| Model | Hyperparameter a | Parameter Value | PCs Included | Mean Accuracy | Std. Dev. |
|---|---|---|---|---|---|
| K-Nearest Neighbors | K | 10 | 10 | 91.6% | 0.3% |
| Support Vector Machine | Kernel | Radial Basis Function | 15 | 94.0% | 0.1% |
| Regularization (C) | 10,000 | ||||
| Gamma | 10 | ||||
| Random Forest | Number of trees | 400 | 16 | 94.3% | 0.1% |
| Maximum depth of trees | 110 | ||||
| Minimum samples for leaf node | 2 | ||||
| Minimum samples to split node | 3 |
a Hyperparameter definitions are provided in the Supplementary Methods.