Table 5 Comparison of the reviewed articles regarding the used development method and model performance (the used symbols SN: sensitivity, SP: specificity, AC: accuracy)
References | Development method/machine learning technique | Model performance |
|---|---|---|
[29] | NB, ABN, and SVM | SN: 80%, 100%, 84%; SP:77%, 100%, 80%; AC: 77%, 100%, 80% |
[30] | CART | SN: 64%, SP: 97%, AC: 80% |
[35] | TC | AC: 97% |
[24] | NB, SVM, and random forests | – |
[33] | DT, ANN, and DT+ANN | AC: 65.18%, 66.25%, 66.25% respectively |
[26] | Bayesian regression | Increase in T ->1.8 ED visit Increase in H ->1 ED visit |
[27] | LR | Coefficient correlation of 0.79 (CO), 0.79 (NO2), 0.93 (PM2.5), and 0.03 (O3) |
[28] | LR | Pearson correlation coefficient: 0.60 |
[34] | GEE | Probability 50% of an asthma attack with low T and high AP |
[31] | PBCAR and PBDT | AC: 86.89% and 87.52%, Recall: 84.12%, and 85.59% respectively |
[23] | Random forest | AC: 80% |
[36] | LR, DT, ANN, SVM, gradient boosting, and random forests | AC, SN, SP, G-mean, precision, and F-measure |
[25] | SVM | AC: 93.55% |