Table 2.
Classification evaluation measures for our classifiers on the GXD dataset using different cross-validation settings
| Classifiers | Precision | Recall | F-measure | Accuracy | Utility-10 | Utility-20 |
|---|---|---|---|---|---|---|
| NB5 | 0.892 (0.005) | 0.957 (0.003) | 0.923 (0.003) | 0.917 (0.004) | 0.876 (0.006) | 0.881 (0.005) |
| RF5 | 0.908 (0.006) | 0.921 (0.005) | 0.915 (0.004) | 0.912 (0.005) | 0.895 (0.007) | 0.899 (0.007) |
| NB10 | 0.891 (0.007) | 0.957 (0.006) | 0.923 (0.004) | 0.917 (0.005) | 0.875 (0.008) | 0.881 (0.008) |
| RF10 | 0.908 (0.008) | 0.922 (0.008) | 0.915 (0.006) | 0.912 (0.007) | 0.894 (0.009) | 0.899 (0.009) |
| RF-H-and-H | 0.905 | 0.925 | 0.915 | 0.913 | 0.896 | 0.900 |
NB denotes Naïve Bayes classifier; RF denotes Random Forest classifier. The suffix 5 indicates using 5 complete runs of 5-fold cross validation; the suffix 10 indicates using 10 complete runs of 10-fold cross validation. H-and-H represents using half of the GXD dataset for training and the other half for testing.