Skip to main content

Table 3 Performance comparison on real datasets

From: A machine learning framework for genotyping the structural variations with copy number variant

 

Coverage

InsertSize

The proposed method

G-Features+M-SVM

G-Features+NB

M-SVM

NB

1

472 ×

216

0.8875

0.9851

0.9777

0.9880

0.9777

2

396 ×

213

0.8917

0.9098

0.9031

0.9469

0.9317

3

394 ×

223

0.9083

0.6265

0.5513

0.7493

0.6766

4

291 ×

175

0.9082

0.6403

0.6839

0.7089

0.8582

5

448 ×

214

0.9000

0.6299

0.7254

0.7444

0.6886

6

460 ×

210

0.9167

0.5881

0.3415

0.6502

0.3589

7

402 ×

202

0.7258

0.4603

0.4632

0.6361

0.6217

8

422 ×

207

0.8917

0.7444

0.7439

0.7660

0.7469

9

369 ×

215

0.9083

0.9234

0.7667

0.9336

0.9669

Average

0.8820

0.7230

0.6840

0.7914

0.7585

Range

0.1909

0.5248

0.6362

0.3519

0.6188

  1. In practice, we adopted the multi-classification support vector machine (M-SVM) as a plus version because the Gindel is a binary classification method based on support vector machine (SVM), it will treat the other three genotypes as classification errors and lead to low accuracy when applied directly