Table 1.
Performance comparison of our doc2vec + RF model with Alguwzizani et al.’s and Barman et al.’s methods using Barman et al.’s dataset.
| Method | SN (%) | SP (%) | ACC (%) | PPV (%) | NPV (%) | MCC | AUC | F1 (%) |
|---|---|---|---|---|---|---|---|---|
| Our model | 81.85 | 76.45 | 79.17 | 77.83 | 80.67 | 0.584 | 0.871 | 79.79 |
| Alguwzizani et al.’s SVMa,b | 73.72 | 83.48 | 78.60 | 81.69 | 76.06 | 0.575 | 0.847 | 77.50 |
| Barman et al.’s SVMa,c,d | 67.00 | 74.00 | 71.00 | 72.00 | NA | 0.440 | 0.730 | 69.41 |
| Barman et al.’s RFa,c,d | 55.66 | 89.08 | 72.41 | 82.26 | NA | 0.480 | 0.760 | 66.39 |
a
The performance was assessed through 5-fold cross-validation.
b
The corresponding values were retrieved from [54].
c
The corresponding values were retrieved from [53].
d
NA means the corresponding parameter is not available. SN: Sensitivity; SP: Specificity; ACC: Accuracy; PPV: Positive Predictive Value (PPV = Precision); NPV: Negative Predictive Value (NPV = TN/(TN + FN)); MCC: Matthews Correlation Coefficient; AUC: the area under the ROC curve; F1 = 2 × (Precision × Recall)/(Precision + Recall).