Table 4: Evaluation results, a) Coefficient; b) Number of people classified, of predictive models on the total dataset and after applying the 10-fold cross validation.
a) |
Method |
Sensitivity |
1-specificity (False Positive Rate) |
F-Measure |
Kappa |
AUROC |
Use full training set |
CART |
0.933 |
0.028 |
0.866 |
0.8477 |
0.956 |
J48 |
0.800 |
0.014 |
0.837 |
0.8178 |
0.956 |
JRip |
0.911 |
0.025 |
0.863 |
0.8449 |
0.944 |
PART |
0.889 |
0.014 |
0.889 |
0.8749 |
0.962 |
Logistic regression |
0.822 |
0.017 |
0.841 |
0.8214 |
0.972 |
10-fold cross validation |
CART |
0.844 |
0.034 |
0.800 |
0.7733 |
0.900 |
J48 |
0.800 |
0.020 |
0.818 |
0.7958 |
0.900 |
JRip |
0.911 |
0.048 |
0.796 |
0.7667 |
0.932 |
PART |
0.778 |
0.028 |
0.778 |
0.7498 |
0.882 |
Logistic regression |
0.822 |
0.025 |
0.813 |
0.7893 |
0.965 |
b) |
CART |
J48 |
JRip |
PART |
LR |
|
Classified as |
|
Real |
COPD |
Non-COPD |
COPD |
Non-COPD |
COPD |
Non-COPD |
COPD |
Non-COPD |
COPD |
Non-COPD |
Use full training set |
COPD |
42 |
3 |
36 |
9 |
41 |
4 |
40 |
5 |
37 |
8 |
Non-COPD |
10 |
347 |
5 |
352 |
9 |
348 |
5 |
352 |
6 |
351 |
10-fold cross validation |
COPD |
38 |
7 |
36 |
9 |
41 |
4 |
35 |
10 |
37 |
8 |
Non-COPD |
12 |
345 |
7 |
350 |
17 |
340 |
10 |
347 |
9 |
348 |