Table 4.
Confusion matrix values for the various classifiers tested.
| TPR | TNR | Accuracy | Precision | Unobscured AGN | Obscured AGN | F1 | |
|---|---|---|---|---|---|---|---|
| (recall) | recall | recall | |||||
| r, 54 visits | 79.5 ± 0.2 | 98.76 ± 0.06 | 95.88 ± 0.05 | 91.8 ± 0.3 | 99.54 ± 0.00 | 48.4 ± 0.6 | 85.2 ± 0.2 |
| r, 33 visits | 79.6 ± 0.9 | 98.23 ± 0.15 | 95.45 ± 0.22 | 88.8 ± 0.9 | 98.6 ± 0.4 | 50.1 ± 1.2 | 83.9 ± 0.8 |
| g, 33 visits | 80.0 ± 0.5 | 98.42 ± 0.15 | 95.67 ± 0.13 | 89.9 ± 0.9 | 98.2 ± 0.3 | 47.8 ± 1.6 | 84.7 ± 0.4 |
| rg, 33 visits | 76.8 ± 0.5 | 98.88 ± 0.07 | 95.59 ± 0.08 | 92.3 ± 0.5 | 98.2 ± 0.0 | 43.1 ± 1.0 | 83.9 ± 0.3 |
| rg + bivar., 33 visits | 73.3 ± 0.7 | 99.06 ± 0.09 | 95.21 ± 0.14 | 93.2 ± 0.6 | 97.9 ± 0.2 | 36.9 ± 1.4 | 82.0 ± 0.6 |
| (g − r)feat, 33 visits | 80.3 ± 0.7 | 98.34 ± 0.09 | 95.64 ± 0.12 | 89.4 ± 0.5 | 99.1 ± 0.2 | 48.7 ± 1.6 | 84.6 ± 0.5 |
| (g − r)mag, 33 visits | 80.0 ± 0.4 | 97.84 ± 0.08 | 95.18 ± 0.09 | 86.7 ± 0.4 | 98.1 ± 0.3 | 50.0 ± 1.5 | 83.2 ± 0.3 |
| r, 33 real, 0 synthetic | 79.5 ± 0.5 | 98.30 ± 0.06 | 95.50 ± 0.09 | 89.2 ± 0.4 | 99.49 ± 0.15 | 48.4 ± 1.4 | 84.1 ± 0.4 |
| r, 29 real, 4 synthetic | 79.7 ± 0.5 | 98.16 ± 0.08 | 95.40 ± 0.09 | 88.4 ± 0.5 | 99.4 ± 0.2 | 48.5 ± 1.0 | 83.8 ± 0.3 |
| r, 25 real, 8 synthetic | 79.5 ± 0.4 | 98.07 ± 0.12 | 95.30 ± 0.09 | 87.9 ± 0.6 | 99.2 ± 0.3 | 48.2 ± 1.5 | 83.5 ± 0.3 |
| r, 21 real, 12 synthetic | 78.9 ± 1.0 | 98.29 ± 0.15 | 95.40 ± 0.17 | 89.0 ± 0.9 | 99.59 ± 0.15 | 45 ± 2 | 83.7 ± 0.6 |
| r, 17 real, 16 synthetic | 79.3 ± 0.4 | 98.23 ± 0.11 | 95.40 ± 0.09 | 88.7 ± 0.6 | 98.9 ± 0.4 | 46.3 ± 1.2 | 83.7 ± 0.3 |
| ks, 25feat, 33 visits | 82.2 ± 0.6 | 98.22 ± 0.08 | 95.82 ± 0.11 | 89.0 ± 0.4 | 98.7 ± 0.2 | 55 ± 2 | 85.5 ± 0.4 |
| ks, 9feat, 33 visits | 85.2 ± 0.3 | 97.75 ± 0.10 | 95.87 ± 0.07 | 86.9 ± 0.5 | 99.0 ± 0.2 | 66.7 ± 1.0 | 86.0 ± 0.2 |
| ks, 8feat, 33 visits | 85.8 ± 0.4 | 97.55 ± 0.11 | 95.80 ± 0.10 | 86.0 ± 0.5 | 99.1 ± 0.3 | 68.1 ± 1.2 | 85.9 ± 0.3 |
| ks, 7feat, 33 visits | 85.4 ± 0.6 | 97.54 ± 0.14 | 95.73 ± 0.15 | 85.9 ± 0.7 | 98.9 ± 0.2 | 67 ± 2 | 85.7 ± 0.5 |
Notes. The table compares true positive ratio (TPR) and true negative ratio (TNR); accuracy, overall precision values, recall for unobscured and obscured AGN, and F1 values obtained in this work are also included, the overall value of the recall being the same as the TPR. All values are to be read as percent values. The percentage errors represent the standard deviation from the mean value derived from a set of ten simulations per classifier. In each simulation, the classifier builds a number of trees as detailed in Appendix A to determine the final classification for each source. What is new in this work is the introduction of additional features (see Sect. 3.2). The last four lines report the corresponding values obtained for four RF classifiers discussed in Sect. 4.2: they are built with the aim of optimizing the identification of obscured AGN in this work.
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.