Table 2. Performance comparison of first-level SVR predictors from PredPPCrys and Crysalis.
| Models | Features# | AUC | MCC | ACC(%) | SPE(%) | SEN(%) | PRE(%) | |
|---|---|---|---|---|---|---|---|---|
| CLF | PredPPCrys | 31 | 0.727 | 0.339 | 67.8 | 62.7 | 71.4 | 73.3 |
| Crysalis | 25 | 0.732 | 0.334 | 66.7 | 68.6 | 65.5 | 76.2 | |
| MF | PredPPCrys | 43 | 0.777 | 0.384 | 70.3 | 69.6 | 71.8 | 50.4 |
| Crysalis | 78 | 0.767 | 0.400 | 70.4 | 68.4 | 75.1 | 49.8 | |
| PF | PredPPCrys | 54 | 0.790 | 0.445 | 73.8 | 70.5 | 75.5 | 83.3 |
| Crysalis | 95 | 0.790 | 0.447 | 74.4 | 69.0 | 77.1 | 83.4 | |
| CF | PredPPCrys | 229 | 0.707 | 0.289 | 62.7 | 74.8 | 58.8 | 87.8 |
| Crysalis | 68 | 0.737 | 0.329 | 70.7 | 73.3 | 63.1 | 85.5 | |
| CRYs | PredPPCrys | 37 | 0.765 | 0.309 | 69.2 | 69.1 | 69.3 | 34.2 |
| Crysalis | 65 | 0.773 | 0.326 | 69.2 | 68.5 | 72.2 | 34.7 |
Performance of all models were evaluated using the benchmark datasets.
#The number of final selected features used for training the first-level SVR predictors.