Skip to main content
. 2022 Jun 7;22(12):4310. doi: 10.3390/s22124310

Table 4.

Performance of ML techniques for ablation study on DXA Model.

Property (No of Features) Evaluation Metric DT MLP RF LR CatBoost XGBoost
Bone Mineral Density (55) Accuracy 0.620 0.682 0.686 0.732 0.710 0.726
Sensitivity 0.617 0.574 0.647 0.678 0.635 0.672
Specificity 0.628 0.795 0.724 0.785 0.784 0.780
Precision 0.624 0.740 0.701 0.758 0.746 0.758
F1-score 0.609 0.639 0.672 0.716 0.685 0.710
MCC 0.250 0.382 0.373 0.466 0.424 0.456
p-value 1.996 × 10−3 1.089 × 10−2 1.664 × 10−3 1.536 × 10−3 6.623 × 10−3 1.332 × 10−3
Body Fat Composition (15) Accuracy 0.720 0.770 0.746 0.742 0.740 0.754
Sensitivity 0.594 0.640 0.697 0.741 0.673 0.723
Specificity 0.832 0.902 0.789 0.749 0.806 0.787
Precision 0.790 0.867 0.770 0.747 0.777 0.767
F1-score 0.669 0.734 0.731 0.741 0.720 0.743
MCC 0.445 0.560 0.489 0.488 0.482 0.508
p-value 1.248 × 10−3 1.175 × 10−3 1.110 × 10−3 1.052 × 10−3 4.975 × 10−3 9.515 × 10−4
Lean Mass (7) Accuracy 0.576 0.652 0.690 0.634 0.668 0.702
Sensitivity 0.556 0.652 0.674 0.613 0.702 0.669
Specificity 0.596 0.664 0.710 0.657 0.638 0.736
Precision 0.582 0.717 0.699 0.641 0.661 0.716
F1-score 0.560 0.650 0.683 0.625 0.678 0.691
MCC 0.156 0.329 0.386 0.270 0.342 0.406
p-value 9.083 × 10−4 6.950 × 10−3 8.326 × 10−4 7.994 × 10−4 3.984 × 10−3 7.402 × 10−4
Area Measurements (45) Accuracy 0.580 0.614 0.600 0.664 0.644 0.598
Sensitivity 0.546 0.528 0.624 0.665 0.660 0.605
Specificity 0.623 0.698 0.584 0.667 0.633 0.597
Precision 0.597 0.636 0.602 0.675 0.644 0.598
F1-score 0.556 0.575 0.607 0.664 0.647 0.598
MCC 0.175 0.230 0.210 0.335 0.295 0.203
p-value 7.138 × 10−4 4.135 × 10−3 6.662 × 10−4 6.447 × 10−4 3.332 × 10−3 6.057 × 10−4
All (122) Accuracy 0.672 0.750 0.748 0.768 0.750 0.774
Sensitivity 0.658 0.656 0.704 0.761 0.694 0.754
Specificity 0.691 0.855 0.797 0.780 0.812 0.800
Precision 0.694 0.827 0.777 0.777 0.789 0.790
F1-score 0.663 0.722 0.736 0.766 0.734 0.768
MCC 0.363 0.526 0.504 0.542 0.511 0.555
p-value 5.879 × 10−4 5.711 × 10−4 5.552 × 10−4 5.402 × 10−4 2.849 × 10−3 5.126 × 10−4