Skip to main content

Table 5 Results: Code Specific SVM Score Thresholding (atc weights). The table shows that SVM score thresholds selected for each code is not an effective strategy compared with thresholds selected for each hierarchy (compare with Table 2).

From: GO for gene documents

Hierarchy

Split

Training FScore

Testing

   

Recall

Precision

FScore

MF

1

0.6221

0.4499

0.3989

0.3852

MF

2

0.615

0.5364

0.4402

0.44351

MF

3

0.6128

0.5295

0.3892

0.4133

MF

4

0.6298

0.5793

0.4394

0.452

MF

5

0.6371

0.5467

0.4264

0.4451

MF

average

0.6234

0.5284

0.4188

0.4278

CC

1

0.5541

0.4679

0.3774

0.3842

CC

2

0.5052

0.5029

0.3435

0.3626

CC

3

0.5131

0.5632

0.3806

0.4239

CC

4

0.5554

0.5134

0.3273

0.3727

CC

5

0.5796

0.5148

0.3875

0.4201

CC

average

0.5415

0.5125

0.3632

0.3927

BP

1

0.4469

0.3994

0.2463

0.2554

BP

2

0.4472

0.4017

0.2727

0.2793

BP

3

0.4378

0.3951

0.2531

0.2589

BP

4

0.4248

0.4309

0.2654

0.2804

BP

5

0.4518

0.3710

0.2434

0.2543

BP

average

0.4417

0.3996

0.2562

0.2657