|
Results: Hierarchy Specific SVM Score Thresholding (ltc weights). The table presents results when a single SVM score threshold is selected for each hierarchy. These results are not significantly different from those for atc weights (compare with Table 2). |
||||||||
| Training |
Testing |
|||||||
| Hierarchy |
Split |
Threshold |
Recall |
Precision |
Fscore |
Recall |
Precision |
Fscore |
|
|
||||||||
| MF |
1 |
-0.83 |
0.5971 |
0.4165 |
0.4640 |
0.6382 |
0.4162 |
0.4735 |
| MF |
2 |
-0.81 |
0.5732 |
0.4194 |
0.4614 |
0.6525 |
0.4537 |
0.5091 |
| MF |
3 |
-0.81 |
0.5687 |
0.4133 |
0.4551 |
0.6522 |
0.4311 |
0.4908 |
| MF |
4 |
-0.83 |
0.6099 |
0.4079 |
0.4634 |
0.6928 |
0.4416 |
0.5138 |
| MF |
5 |
-0.83 |
0.6107 |
0.4148 |
0.4672 |
0.6880 |
0.4036 |
0.4822 |
| MF |
Average |
na |
0.5919 |
0.4144 |
0.4622 |
0.6647 |
0.4292 |
0.4939 |
|
|
||||||||
| CC |
1 |
-0.83 |
0.5426 |
0.3135 |
0.3772 |
0.5383 |
0.3248 |
0.3810 |
| CC |
2 |
-0.81 |
0.5075 |
0.3248 |
0.3765 |
0.5422 |
0.3437 |
0.3971 |
| CC |
3 |
-0.83 |
0.5194 |
0.3136 |
0.3685 |
0.6488 |
0.4075 |
0.4763 |
| CC |
4 |
-0.85 |
0.5606 |
0.3080 |
0.3779 |
0.6216 |
0.3161 |
0.3929 |
| CC |
5 |
-0.84 |
0.5689 |
0.3284 |
0.3956 |
0.6137 |
0.3136 |
0.3993 |
| CC |
Average |
na |
0.5398 |
0.3177 |
0.3791 |
0.5929 |
0.3411 |
0.4093 |
|
|
||||||||
| BP |
1 |
-0.82 |
0.3773 |
0.2596 |
0.2881 |
0.4204 |
0.2894 |
0.3163 |
| BP |
2 |
-0.84 |
0.4124 |
0.2595 |
0.2965 |
0.4912 |
0.2898 |
0.3423 |
| BP |
3 |
-0.86 |
0.4405 |
0.2522 |
0.2983 |
0.5366 |
0.2737 |
0.3395 |
| BP |
4 |
-0.85 |
0.3935 |
0.2344 |
0.2713 |
0.5211 |
0.2856 |
0.3497 |
| BP |
5 |
-0.85 |
0.4472 |
0.2691 |
0.3131 |
0.4680 |
0.2466 |
0.3035 |
| BP |
Average |
na |
0.4142 |
0.2550 |
0.2935 |
0.4875 |
0.2770 |
0.3303 |
Srinivasan and Qiu BMC Bioinformatics 2007 8(Suppl 9):S3 doi:10.1186/1471-2105-8-S9-S3 |
||||||||