Skip to main content

Table 12 Comparison of the performance K-Nearest Neighbor classifier (KNN) using different values of k {1, 3, 5, 10} neighbors and using different distance functions; Euclidean distance, Manhattan distance and Minkowski distance using sampling

From: Comparison of machine learning techniques to predict all-cause mortality using fitness data: the Henry ford exercIse testing (FIT) project

  Euclidean distance Manhattan Distance Minkowski Distance
K = 1 K = 3 K = 5 K = 10 K = 1 K = 3 K = 5 K = 10 K = 1 K = 3 K = 5 K = 10
Sensitivity 78.43% 65.61% 64.17% 50.00% 78.29% 65.66% 65.68% 61.23% 78.43% 65.61% 64.17% 59.23%
Specificity 96.98% 91.74% 90.53% 89.84% 97.05% 91.80% 90.60% 89.91% 96.98% 91.74% 90.53% 89.84%
Precision 77.18% 33.64% 22.32% 11.50% 77.73% 34.16% 22.94% 16.44% 77.18% 33.64% 22.32% 15.89%
F-score 77.80% 44.47% 33.12% 18.70% 78.01% 44.94% 34.01% 25.91% 77.80% 44.47% 33.12% 25.05%
RMSE 0.23 0.27 0.28 0.29 0.23 0.27 0.28 0.29 0.23 0.27 0.28 0.29
AUC 0.88 0.86 0.85 0.84 0.87 0.86 0.85 0.84 0.87 0.86 0.85 0.84
  1. The results show that the value 1 for the K parameter achieves the highest AUC (0.88) using Euclidean distance