Table 2. Summary of the performance of PCM models.
Classification method | Maximum height of signature | Cross-validation | External prediction | ||
accuracy (%) | AUC | accuracy | AUC | ||
Support Vector Machine | 1 | 83.73 | 0.905 | 82.52 | 0.883 |
2 | 85.13 | 0.920 | 87.04 | 0.922 | |
3* | 85.68 | 0.923 | 88.33 | 0.940 | |
Random Forest | 1 | 83.03 | 0.900 | 87.53 | 0.933 |
2 | 84.35 | 0.915 | 88.32 | 0.941 | |
3 | 84.16 | 0.918 | 88.63 | 0.946 | |
k-Nearest Neighbors | 1 | 78.97 | 0.865 | 80.38 | 0.866 |
2 | 79.56 | 0.870 | 80.10 | 0.868 | |
3 | 78.48 | 0.860 | 79.54 | 0.865 |
Model implemented in Bioclipse Decision Support at www.cyp450model.org.