A Machine Learning Approach to the Interpretation of Cardiopulmonary Exercise Tests: Development and Validation

Table 1

Outcomes (probability estimates (%)) of the SVM multilabel crossvalidation procedure of the learning stage.

Conventional clinical diagnosis	Model cross-validation		SVM classification (%)
	Sample’s splits	No. of iterations	CHF	COPD	Healthy
CHF	Leave-one-out	150
COPD
Healthy
CHF	80% (training) 20% (validation)	300
COPD
Healthy
CHF	70% (training) 30% (validation)	675
COPD
Healthy
CHF	50% (training) 50% (validation)	1875
COPD
Healthy

Data presented as . CHF: chronic heart failure; COPD: chronic obstructive pulmonary disease; Healthy: healthy normal participants; SVM: support vector machine, SD: standard deviation. Bold numbers denote (%) probability estimates of the respective group.