Table 4.
Comparison of the proposed method with the most relevant works.
| Author & References | Classifiers | Accuracy (%) | Precision (%) | AUC (%) | Limitations |
|---|---|---|---|---|---|
| G. Sailasya et al. [18] | LR | 78.00 | 77.50 | – | a) Used limited ML algorithms that were not robust for accurate stroke prediction. |
| DT | 66.00 | 77.50 | – | ||
| RF | 73.00 | 72.00 | – | ||
| K-NN | 80.00 | 77.40 | – | ||
| SVM | 80.00 | 78.60 | – | ||
| NB | 82.00 | 79.20 | – | ||
| Krishna et al. [13] | RF | 90.36 | 88.00 | – | b) Lower accuracy and AUC score which is not enough for stroke prediction. |
| LR | 80.18 | 79.00 | – | ||
| SVM | 80.18 | 79.00 | – | ||
| KNN | 86.74 | 83.00 | – | ||
| NB | 76.03 | 74.00 | – | ||
| XGB | 89.02 | 88.00 | – | ||
| Kokkotis et al. [20] | LR | 73.52 | – | 83.30 | c) Focused on specific ages, removed prevailing missing dataset values. |
| RF | 71.19 | – | 81.24 | ||
| XGboost | 72.58 | – | 82.50 | ||
| KNN | 69.16 | – | 79.35 | ||
| SVM | 71.28 | – | 82.85 | ||
| MLP | 70.85 | – | 82.14 | ||
| PCA-FA | LR | 80.48 | 75.51 | 88.46 | |
| RF | 92.55 | 90.53 | 98.15 | ||
| KNN | 89.54 | 82.16 | 94.16 | ||
| SVM | 86.38 | 80.72 | 93.35 | ||
| GB | 86.38 | 82.03 | 92.75 | ||
| XGB | 91.52 | 87.58 | 97.37 |