Performance evaluation of k-NN model.

Résumé fourni par la source

(a) Classification boundaries generated using k-nearest neighbors (k-NN) to distinguish samples with low (blue) and high (red) SARS-CoV-2 seroreactivity. The k-NN model was trained on the first three principal components of PCA-transformed serology data from healthy children sampled post-pandemic. Data points are colored based on hierarchical clustering classification (as in Fig 2b), while background shading indicates k-NN decision boundaries. (b) Performance of the k-NN model in classifying seroreactivity phenotypes among healthy post-pandemic samples when evaluated by 5-fold cross-validation. The model achieved 94% accuracy, 96% sensitivity, and 91% specificity. These values are shown relative to the cross-validated accuracy from 1000 label-permuted datasets. P-values were derived from the tail probability of the null distribution. PC: principal component. (PDF)

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.