Table 1 Comparison between the proposed system and different machine learning techniques.
| Â | Classifier | Metrics | Class evaluation | Overall evaluation | |||
|---|---|---|---|---|---|---|---|
Level 0 | Level 1 | Level 2 | Overall accuracy (%) | Kohen Kappa (%) | |||
4-fold | Random forest29 | Sensitivity (%) | \(90.79 \pm 8.89\) | \(57.80 \pm 6.24\) | \(62.39 \pm 10.82\) | \(76.81 \pm 3.84\) | \(61.04 \pm 6.06\) |
Specificity (%) | \(71.74 \pm 8.13\) | \(95.18 \pm 3.77\) | \(91.43 \pm 5.4\) | ||||
Decision trees30 | Sensitivity (%) | \(83.27 \pm 7.98\) | \(51.46 \pm 9.29\) | \(46.04 \pm 17.58\) | \(67.53 \pm 7.04\) | \(45.88 \pm 11.25\) | |
Specificity (%) | \(70.8 \pm 6.19\) | \(86.55 \pm 7.74\) | \(88.63 \pm 5.83\) | ||||
Naive Bayes31 | Sensitivity (%) | \(88.29 \pm 5.07\) | \(67.28 \pm 7.57\) | \(65.17 \pm 13.4\) | \(78.23 \pm 4.53\) | \(63.98 \pm 7.29\) | |
Specificity (%) | \(78.49 \pm 13.14\) | \(93.89 \pm 2.19\) | \(90.52 \pm 3.38\) | ||||
SVM32 | Sensitivity (%) | \(90.11 \pm 5.6\) | \(63.29 \pm 5.72\) | \(67.87 \pm 14.51\) | \(78.94 \pm 1.4\) | \(64.77 \pm 1.62\) | |
Specificity (%) | \(76.67 \pm 8.09\) | \(95.14 \pm 2.46\) | \(91.32 \pm 2.22\) | ||||
KNN33 | Sensitivity (%) | \(90.79 \pm 7.00\) | \(68.23 \pm 17.28\) | \(67.97 \pm 17.59\) | \(78.91 \pm 1.89\) | \(64.5 \pm 3.08\) | |
Specificity (%) | \(72.4 \pm 9.31\) | \(95.54 \pm 2.28\) | \(93.42 \pm 3.89\) | ||||
AdaBoost34 | Sensitivity (%) | \(91.36 \pm 7.95\) | \(62.07 \pm 8.79\) | \(49.93 \pm 15.53\) | \(75.18 \pm 4.21\) | \(58.62 \pm 7.13\) | |
Specificity (%) | \(71.12 \pm 9.5\) | \(92.26 \pm 2.84\) | \(92.65 \pm 4.15\) | ||||
Proposed system | Sensitivity (%) | \({\textbf {93.27}} \pm {\textbf {3.68}}\) | \({\textbf {92}} \pm {\textbf {0}}\) | \({\textbf {93}} \pm {\textbf {2}}\) | \({\textbf {92.77}} \pm {\textbf {1.7}}\) | \({\textbf {89.18}} \pm {\textbf {2.52}}\) | |
Specificity (%) | \({\textbf {95}} \pm {\textbf {1.15}}\) | \({\textbf {97.55}} \pm {\textbf {1.88}}\) | \({\textbf {96.57}} \pm {\textbf {0.98}}\) | ||||
5-fold | Random forest29 | Sensitivity (%) | \(92.47 \pm 6.67\) | \(63.77 \pm 11.53\) | \(64.23 \pm 13.48\) | \(79.46 \pm 2.36\) | \(65.60 \pm 4.62\) |
Specificity (%) | \(73.29 \pm 10.11\) | \(95.95 \pm 2.78\) | \(93.17 \pm 5.38\) | ||||
Decision trees30 | Sensitivity (%) | \(78.93 \pm 8.95\) | \(60.27 \pm 13.82\) | \(43.90 \pm 17.26\) | \(66.61 \pm 5.13\) | \(45.90 \pm 9.16\) | |
Specificity (%) | \(71.43 \pm 9.86\) | \(86.14 \pm 5.89\) | \(87.41 \pm 9.39\) | ||||
Naive Bayes31 | Sensitivity (%) | \(89.40 \pm 3.95\) | \(66.20 \pm 10.97\) | \(62.98 \pm 15.64\) | \(78.08 \pm 3.56\) | \(63.57 \pm 5.87\) | |
Specificity (%) | \(77.96 \pm 13.38\) | \(93.48 \pm 3.21\) | \(91.00 \pm 3.77\) | ||||
SVM32 | Sensitivity (%) | \(92.40 \pm 3.29\) | \(69.36 \pm 11.85\) | \(63.27 \pm 9.88\) | \(80.43 \pm 2.17\) | \(67.37 \pm 4.33\) | |
Specificity (%) | \(77.5 \pm 8.55\) | \(95.18 \pm 3.03\) | \(92.68 \pm 3.26\) | ||||
KNN33 | Sensitivity (%) | \(94.72 \pm 2.43\) | \(63.77 \pm 11.53\) | \(61.73 \pm 12.07\) | \(80.03 \pm 1.54\) | \(66.52 \pm 3.04\) | |
Specificity (%) | \(71.33 \pm 9.41\) | \(96.34 \pm 0.80\) | \(94.71 \pm 2.95\) | ||||
AdaBoost34 | Sensitivity (%) | \(91.88 \pm 5.56\) | \(62.11 \pm 11.61\) | \(49.81 \pm 15.62\) | \(75.37 \pm 3.51\) | \(59.12 \pm 5.85\) | |
Specificity (%) | \(70.01 \pm 14.42\) | \(92.72 \pm 3.27\) | \(92.43 \pm 6.25\) | ||||
Proposed system | Sensitivity (%) | \({\textbf {92.38}} \pm {\textbf {2.61}}\) | \({\textbf {93.05}} \pm {\textbf {2.78}}\) | \({\textbf {96}} \pm {\textbf {2.24}}\) | \({\textbf {93.79}} \pm {\textbf {1.37}}\) | \({\textbf {90.68}} \pm {\textbf {2.05}}\) | |
Specificity (%) | \({\textbf {96.52}} \pm {\textbf {1.34}}\) | \({\textbf {97.56}} \pm {\textbf {1.73}}\) | \({\textbf {96.6}} \pm {\textbf {2.19}}\) | ||||
10-fold | Random forest29 | Sensitivity (%) | \(91.54 \pm 18.57\) | \(67.03 \pm 7.02\) | \(65.39 \pm 16.86\) | \(79.95 \pm 4.94\) | \(66.5 \pm 9.01\) |
Specificity (%) | \(77.84 \pm 6.51\) | \(94.91 \pm 14.61\) | \(92.16 \pm 3.77\) | ||||
Decision trees30 | Sensitivity (%) | \(83.55 \pm 19.80\) | \(61.42 \pm 9.31\) | \(51.98 \pm 22.75\) | \(71.08 \pm 6.97\) | \(52.75 \pm 12.43\) | |
Specificity (%) | \(71.12 \pm 6.86\) | \(91.14 \pm 15.09\) | \(88.53 \pm 6.30\) | ||||
Naive Bayes31 | Sensitivity (%) | \(90.28 \pm 21.38\) | \(68.46 \pm 6.91\) | \(63.69 \pm 14.36\) | \(79.26 \pm 3.65\) | \(65.62 \pm 6.94\) | |
Specificity (%) | \(78.51 \pm 5.56\) | \(93.74 \pm 14.52\) | \(92.07 \pm 5.73\) | ||||
SVM32 | Sensitivity (%) | \(92.54 \pm 21.83\) | \(67.71 \pm 5.73\) | \(65.25 \pm 18.28\) | \(80.63 \pm 5.77\) | \(67.57 \pm 11.09\) | |
Specificity (%) | \(76.75 \pm 4.62\) | \(95.83 \pm 17.16\) | \(92.73 \pm 3.72\) | ||||
KNN33 | Sensitivity (%) | \(94.90 \pm 24.36\) | \(64.67 \pm 4.46\) | \(65.43 \pm 10.29\) | \(81.14 \pm 4.82\) | \(68.45 \pm 8.41\) | |
Specificity (%) | \(73.56 \pm 4.69\) | \(96.81 \pm 11.73\) | \(94.52 \pm 2.37\) | ||||
AdaBoost34 | Sensitivity (%) | \(86.04 \pm 21.96\) | \(64.53 \pm 19.02\) | \(60.69 \pm 18.44\) | \(77.13 \pm 5.54\) | \(61.89 \pm 10.16\) | |
Specificity (%) | \(71.52 \pm 3.74\) | \(92.2 \pm 17.29\) | \(94.94 \pm 4.14\) | ||||
Proposed system | Sensitivity (%) | \({\textbf {96.27}} \pm {\textbf {4.82}}\) | \({\textbf {99}} \pm {\textbf {3.16}}\) | \({\textbf {98}} \pm {\textbf {4.26}}\) | \({\textbf {97.72}} \pm {\textbf {1.57}}\) | \({\textbf {96.58}} \pm {\textbf {2.36}}\) | |
Specificity (%) | \({\textbf {99.5}} \pm {\textbf {1.58}}\) | \({\textbf {98.55}} \pm {\textbf {2.34}}\) | \({\textbf {98.55}} \pm {\textbf {2.34}}\) | ||||
