Skip to main content

Table 16 Comparison of the performance of the different classification models without using the SMOTE sampling method

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

 

DT

SVM

ANN

BC

BN

KNN

RF

Sensitivity

54.43%

39.78%

52.65%

37.71%

57.09%

46.78%

59.09%

Specificity

90.51%

87.65%

91.37%

93.23%

90.71%

89.31%

90.21%

Precision

22.80%

18.03%

31.39%

51.32%

24.57%

11.12%

19.52%

F-score

32.14%

24.81%

39.33%

43.47%

34.35%

17.97%

29.35%

RMSE

0.3

0.47

0.29

0.34

0.34

0.3

0.29

AUC

0.73

0.59

0.80

0.82

0.83

0.74

0.82

  1. The models are: Decision Tree (DT), Support Vector Machine (SVM), Artificial Neural Networks (ANN), Naïve Bayesian Classifier (BC), Bayesian Network (BN), K-Nearest Neighbor (KNN) and Random Forest (RF). The results of this experiment show that BN achieves the highest AUC (0.83). The BC model achieves the highest precision (51.32%) and the highest specificity (93.32%)