As can be seen in Table V, the set of all feature (FS0)
shows a more positive result compared with other sets in
almost of all classification algorithms. Two out of six classification
algorithms gained the best result on both specificity
and sensitivity scores with the set of all features. With regards
to specificity score, FS0 is the most suitable feature-set
for SVM, LR and DT, while RF, J48 and DT perform the
best on the 6-feature-set (FS2). In terms of sensitivity, FS0 is
considerably higher than other sets as it is the best choice for
4 algorithms.
With focus on the best algorithm (Logistic Regression)
that is analysed above, FS0 is the best choice in both specificity
and sensitivity, at 91.4% and 90.2%, respectively.