Table 4 Characteristics of the cross validation experiments for the Mycobacterium tuberculosis AdaBoost classifiers.

From: Antimicrobial Resistance Prediction in PATRIC and RAST

 

Available Genomes

Genomes used per trial *

Classifier statistics

Antibiotic

RES

SUS

Test set

Training set

AUC

F1 plot point

F1 Score

Accuracy at F1 point

Accuracy at point zero

Ethambutol

250

250

25

225

0.715

0.435

0.704

0.588

0.668

Ethionamide

173

250

17

153

0.812

−0.136

0.766

0.768

0.771

Isoniazid

250

250

25

225

0.911

−0.085

0.872

0.880

0.882

Kanamycin

188

250

18

162

0.898

0.137

0.871

0.883

0.872

Ofloxacin

239

250

23

207

0.833

−0.022

0.761

0.793

0.791

Rifampicin

250

250

25

225

0.932

−0.410

0.870

0.864

0.858

Streptomycin

250

250

25

225

0.795

−0.485

0.722

0.642

0.712

Combined Set

83

139

8

72

0.969

−0.577

0.950

0.950

0.928

  1. *For each round of cross validation the depicted set size was chosen for the susceptible and resistant genomes.
Back to article page