Table 3 Model metrics for each iteration.
From: A machine learning approach for non-invasive PCOS diagnosis from ultrasound and clinical features
Feature set | Training AUC | Training precision | Training F1 score | Training accuracy | Test AUC | Test precision | Test F1 score | Test accuracy |
|---|---|---|---|---|---|---|---|---|
1 st iteration: Only Ultrasound Features | 0.9636 | 0.8855 | 0.8406 | 0.8981 | 0.8312 | 0.7037 | 0.6441 | 0.8073 |
2nd iteration: Ultrasound & Biomarker Features | 0.9986 | 0.9789 | 0.9686 | 0.9792 | 0.8364 | 0.6452 | 0.6349 | 0.789 |
3rd iteration: Clinical & Biomarker Features | 0.9997 | 0.9863 | 0.9897 | 0.9931 | 0.8482 | 0.6364 | 0.5185 | 0.7615 |
4th iteration: Ultrasound & Clinical Features | 0.9999 | 1 | 0.9895 | 0.9931 | 0.9545 | 0.8571 | 0.8 | 0.8899 |
5th iteration: Optimized Ultrasound & Clinical Features | 1 | 1 | 1 | 1 | 0.9852 | 0.9583 | 0.9388 | 0.9384 |
6th iteration: Feature Selection | 1 | 1 | 1 | 1 | 0.9947 | 0.9553 | 0.9553 | 0.9553 |
