Table 2.
Summary of machine learning results expressed as percentage ratios over ElectroShape 5D. A value of 100% indicates that the same performance as ElectroShape 5D was obtained.
| LEC | GMM | Isolation Forest | ANN | ANN |
|---|---|---|---|---|
| (Hidden layer = 100) | (Hidden layer = 500) | |||
| Mean EF1%(± std) | 291%(± 162%) | 191%(± 084%) | 256%(± 107%) | 333%(± 130%) |
| Max EF1% | 829% | 450% | 491% | 618% |
| Min EF1% | 0% | 0% | 57% | 121% |
| Mean AUC(± std) | 133%(± 17%) | 126%(± 15%) | 139%(± 17%) | 144%(± 21%) |
| Max AUC | 171% | 155% | 175% | 179% |
| Min AUC | 104% | 99% | 104% | 105% |
| Full Conformers | ||||
| Mean EF1%(± std) | 430%(± 223%) | 211%(± 90%) | 256%(± 129%) | 328%(± 149%) |
| Max EF1% | 941% | 403% | 565% | 636% |
| Min EF1% | 107% | 0% | 82% | 30% |
| Mean AUC(± std) | 137%(± 19%) | 124%(± 14%) | 136%(± 20%) | 143%(± 20%) |
| Max AUC | 173% | 153% | 173% | 177% |
| Min AUC | 105% | 99% | 103% | 104% |