Table 2.
The performance of diatom composition based models developed using different machine learning techniques.
| Machine learning techniques (parameters) | Main dataset |
Validation dataset |
||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Sen | Spc | Acc | MCC | AUROC | Sen | Spc | Acc | MCC | AUROC | |
| SVM (g = 0.05, c = 15, j = 2) | 89.71 | 86.85 | 88.29 | 0.77 | 0.93 | 90.53 | 81.72 | 86.17 | 0.73 | 0.92 |
| Random Forest (Ntree = 150) | 94.20 | 85.23 | 89.75 | 0.80 | 0.96 | 92.11 | 82.80 | 87.50 | 0.75 | 0.93 |
| SMO (g = 0.1, c = 5) | 88.79 | 87.92 | 88.36 | 0.77 | 0.88 | 88.95 | 83.33 | 86.17 | 0.72 | 0.86 |
| J48 (c = 0.25, m = 1) | 89.71 | 83.22 | 86.49 | 0.73 | 0.88 | 86.84 | 83.87 | 85.37 | 0.71 | 0.86 |
| Naive Bayes (Default) | 87.86 | 63.09 | 75.58 | 0.53 | 0.74 | 87.37 | 62.37 | 75.00 | 0.51 | 0.74 |
Sen, Sensitivity; Spc, Specificity; Acc, Accuracy; MCC, Matthew’s Correlation Coefficient; AUROC, Area Under the Receiver Operating Characteristic curve.