Table 4 The performance metrics of ML-based algorithms with their confidence interval for predicting malnutrition of women in Bangladesh using BDHS, 2017-18 data
| Â | Training dataset | Test dataset | ||||
|---|---|---|---|---|---|---|
(a) Naïve Bayes | ||||||
Confusion Matrix | Â Category | No | Yes | Â Category | No | Yes |
No | 4890 | 2127 | No | 1610 | 704 | |
Yes | 3066 | 2683 | Yes | 1042 | 900 | |
Sensitivity | 0.615 (0.604, 0.626) | 0.607 (0.588, 0.626) | ||||
Specificity | 0.558 (0.544, 0.572) | 0.561 (0.537, 0.585) | ||||
Positive Predictive Value | 0.697 (0.686, 0.708) | 0.696 (0.677, 0.715) | ||||
Negative Predictive Value | 0.467 (0.454, 0.480) | 0.463 (0.441, 0.485) | ||||
Accuracy | 0.593 (0.584, 0.602) | 0.590 (0.575, 0.605) | ||||
\(F_1\)Score | 0.653 | 0.648 | ||||
Kappa | 0.166 | 0.162 | ||||
(b) CART | ||||||
Confusion Matrix | Â Category | No | Yes | Â Category | No | Yes |
No | 5622 | 1395 | No | 1841 | 473 | |
Yes | 3642 | 2107 | Yes | 1270 | 672 | |
Sensitivity | 0.607 (0.597, 0.617) | 0.592 (0.575, 0.609) | ||||
Specificity | 0.602 (0.586, 0.618) | 0.587 (0.558, 0.616) | ||||
Positive Predictive Value | 0.801 (0.792, 0.810) | 0.796 (0.780, 0.812) | ||||
Negative Predictive Value | 0.366 (0.354, 0.378) | 0.346 (0.325, 0.367) | ||||
Accuracy | 0.605 (0.597, 0.613) | 0.591 (0.575, 0.605) | ||||
\(F_1\)Score | 0.691 | 0.679 | ||||
Kappa | 0.174 | 0.147 | ||||
(c) C5.0 | ||||||
Confusion Matrix | Â Category | No | Yes | Â Category | No | Yes |
No | 5789 | 1228 | No | 1904 | 410 | |
Yes | 3903 | 1846 | Yes | 1331 | 611 | |
Sensitivity | 0.597 (0.587, 0.607) | 0.589 (0.572, 0.606) | ||||
Specificity | 0.601 (0.584, 0.618) | 0.598 (0.568, 0.628) | ||||
Positive Predictive Value | 0.825 (0.816, 0.834) | 0.823 (0.807, 0.839) | ||||
Negative Predictive Value | 0.321 (0.309, 0.333) | 0.315 (0.294, 0.336) | ||||
Accuracy | 0.598 (0.589, 0.607) | 0.591 (0.576, 0.606) | ||||
\(F_1\)Score | 0.693 | 0.686 | ||||
Kappa | 0.153 | 0.143 | ||||
(d) Logistic Regression | ||||||
Confusion Matrix | Â Category | No | Yes | Â Category | No | Yes |
No | 6729 | 288 | No | 1799 | 515 | |
Yes | 5094 | 655 | Yes | 1215 | 727 | |
Sensitivity | 0.569 (0.560, 0.578) | 0.597 (0.579, 0.615) | ||||
Specificity | 0.695 (0.666, 0.724) | 0.585 (0.558, 0.612) | ||||
Positive Predictive Value | 0.959 (0.954, 0.964) | 0.777 (0.760, 0.794) | ||||
Negative Predictive Value | 0.114 (0.106, 0.122) | 0.374 (0.352, 0.396) | ||||
Accuracy | 0.578 (0.569, 0.587) | 0.594 (0.579, 0.609) | ||||
\(F_1\)Score | 0.714 | 0.675 | ||||
Kappa | 0.079 | 0.153 | ||||
(e) Random Forest | ||||||
Confusion Matrix | Â Category | No | Yes | Â Category | No | Yes |
No | 5735 | 1282 | No | 1902 | 412 | |
Yes | 3824 | 1925 | Yes | 1284 | 658 | |
Sensitivity | 0.600 (0.590, 0.610) | 0.597 (0.580, 0.614) | ||||
Specificity | 0.600 (0.583, 0.617) | 0.615 (0.586, 0.644) | ||||
Positive Predictive Value | 0.817 (0.808, 0.826) | 0.822 (0.806, 0.838) | ||||
Negative Predictive Value | 0.335 (0.323, 0.347) | 0.339 (0.318, 0.360) | ||||
Accuracy | 0.600 (0.592, 0.608) | 0.602 (0.587, 0.617) | ||||
\(F_1\)Score | 0.692 | 0.692 | ||||
Kappa | 0.159 | 0.167 | ||||
(f) Gradient Boosting | ||||||
Confusion Matrix | Category | No | Yes |  Category | No | Yes |
No | 5442 | 1575 | No | 1796 | 518 | |
Yes | 3517 | 2232 | Yes | 1203 | 739 | |
Sensitivity | 0.607 (0.597, 0.617) | 0.599 (0.581, 0.617) | ||||
Specificity | 0.586 (0.570, 0.602) | 0.588 (0.561, 0.615) | ||||
Positive Predictive Value | 0.776 (0.766, 0.786) | 0.776 (0.759, 0.793) | ||||
Negative Predictive Value | 0.388 (0.375, 0.401) | 0.381 (0.359, 0.403) | ||||
Accuracy | 0.601 (0.593, 0.609) | 0.596 (0.581, 0.611) | ||||
\(F_1\)Score | 0.681 | 0.676 | ||||
Kappa | 0.169 | 0.161 | ||||