Skip to main content

Table 13 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 without 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 28.06% 38.19% 42.44% 46.78% 28.54% 38.21% 42.96% 47.59% 28.06% 38.19% 42.44% 28.06%
Specificity 90.24% 89.88% 89.50% 89.31% 90.28% 89.87% 89.49% 89.31% 90.24% 89.88% 89.50% 90.24%
Precision 25.36% 18.37% 13.64% 11.12% 25.62% 18.18% 13.42% 11.12% 25.36% 18.37% 13.64% 25.36%
F-score 26.64% 24.81% 20.64% 17.97% 27.00% 24.64% 20.45% 18.03% 26.64% 24.81% 20.64% 26.64%
RMSE 0.4 0.33 0.32 0.3 0.4 0.33 0.32 0.31 0.4 0.33 0.32 0.4
AUC 0.58 0.66 0.7 0.74 0.59 0.67 0.7 0.74 0.58 0.66 0.7 0.58
  1. The results show that the value 10 for the K parameter achieves the highest AUC (0.74) using Euclidean distance