The final prediction model classified whether a cow was within 5 hours of calving. As shown in the figure, the model achieved an average Precision of 0.743, indicating a high proportion of correct alerts. The average Recall was 0.626, showing that the model successfully identified many, though not all, imminent calving cases.
This balance between precision and recall demonstrates the model’s practical value for early warning, with Tail Raising identified as the most informative behavior.