Table 4. The quantitative results for accuracy using different regularization parameters.
For each dataset, 80% was used to train a prediction model and the remaining 20% for testing.
| Classifier | Dataset | Regularization Parameter | Accuracy |
|---|---|---|---|
| SVM | Abstracts | L2 (Default) | 94.63% |
| SVM | Abstracts | L1 | 91.07% |
| SVM | Abstracts | None | 89.72% |
| Logistic Regression | Abstracts | L2 (Default) | 92.19% |
| Logistic Regression | Abstracts | L1 | 90.61% |
| Logistic Regression | Abstracts | None | 88.54% |
| SVM | Full-text Articles I | L2 (Default) | 94.47% |
| SVM | Full-text Articles I | L1 | 90.33% |
| SVM | Full-text Articles I | None | 88.51% |
| Logistic Regression | Full-text Articles I | L2 (Default) | 91.05% |
| Logistic Regression | Full-text Articles I | L1 | 88.19% |
| Logistic Regression | Full-text Articles I | None | 87.04% |
| SVM | Full-text Articles II | L2 (Default) | 93.81% |
| SVM | Full-text Articles II | L1 | 90.16% |
| SVM | Full-text Articles II | None | 87.94% |
| Logistic Regression | Full-text Articles II | L2 (Default) | 90.57% |
| Logistic Regression | Full-text Articles II | L1 | 87.63% |
| Logistic Regression | Full-text Articles II | None | 86.71% |