Table 2.
Performance of different classification models.
| Both groups | National tennis players | International tennis players | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AUC | CA | Se | Sp | AUC | CA | Se | Sp | AUC | CA | Se | Sp | |
| Server | 0.686 | 0.215 | 0.201 | 0.214 | 0.834 | 0.453 | 0.577 | 0.526 | 0.901 | 0.575 | 0.575 | 0.575 |
| Returning player | 0.953 | 0.664 | 0.735 | 0.664 | 0.970 | 0.780 | 0.801 | 0.780 | 0.883 | 0.642 | 0.710 | 0.642 |
| Return category | 0.667 | 0.523 | 0.595 | 0.523 | 0.717 | 0.581 | 0.621 | 0.581 | 0.753 | 0.549 | 0.583 | 0.549 |
AUC, area under the curve; CA, classification accuracy; Se, sensitivity; Sp, specificity are presented for classifying individual server, returning player, and return category for all returns combined and for returns made by national and international tennis players separately.