Table 3.
Comparison of the graph classification using unlabeled and labeled graphs
| Classifier | Attributes | Acc. | Sens. | Spec. | Prec. | F-Measure |
|---|---|---|---|---|---|---|
| Random Forest | 7U | 63.2 | 65.2 | 60.9 | 67.0 | 66.1 |
| σ | 0.77 | 1.02 | 0.83 | 0.67 | 0.79 | |
| Random Forest | 5U + 2L | 64.0 | 66.5 | 60.9 | 67.5 | 67.0 |
| σ | 0.88 | 1.46 | 1.87 | 0.97 | 1.15 | |
| SVM | 7U | 63.0 | 83.3 | 38.2 | 62.2 | 71.2 |
| σ | 1.23 | 2.66 | 4.92 | 1.32 | 1.67 | |
| SVM | 5U + 2L | 65.0 | 79.3 | 47.7 | 64.9 | 71.4 |
| σ | 0.88 | 1.07 | 2.37 | 0.90 | 0.97 | |
The table contains the results of the graph classification applying the information-theoretic descriptors for vertexand edge-labeled graphs. Here, U indicates the usage of a measure only defined for unlabeled graphs and L indicates the usage of a measure for vertex- and edge-labeled graphs, respectively. σ denotes the standart deviation of the corresponding means.