DeVries Z, Hoda M, Rivers C, et al. - In this retrospective review of a prospective cohort study of individuals enrolled between 2004 and 2017 with the complete neurological examination and Functional Independence Measure outcome data at ≥ 1-year follow-up or who could walk at discharge, experts evaluated the performance of both an unsupervised machine learning algorithms (MLA) and logistic regression (LR) model with complete admission neurological data against the van Middendorp et al. (2011) and Hicks models and compared the exactitude of the area under the receiver operators curve (AUROC) and the F1-score to ascertain which method was better for the evaluation of diagnostic performance of prediction models on large-scale datasets. Between the use of an unsupervised MLA with complete admission neurological information in comparison with the formerly authorized standards, no clinically important variations were discovered, nevertheless, when contrasting the performance of the AUROC and F1-score, the incorrect prognostic performance was exhibited by the AUROC when there was an imbalance towards a greater amount of false negatives. Significantly, the F1-score did not accede to this imbalance. When assessing the performance of prediction models, As AUROC has been used as the standard, consideration as to whether this is the most relevant method is guaranteed.