wait a moment

Spectroscopic approach to capillary-alveolar membrane damage: APPENDIX (2)

Lung injury

The amount of HES in the fluids varied from 0.03 to 3 |ig/mL. For cross-validation of the training set, the leave-one-out method was applied, which is commonly used for evaluating the prediction ability of training processes. It is performed by sequentially removing a single spectrum from the training set, then training the classifier on the remaining spectra. Prediction is made for that single test spectrum. This process is repeated for each spectrum in the training set, and the classifier returned to the training set after each test so that the set remains the same size. ventolin inhalers
The purpose of this cross-validation test is to improve the probability that the training set truly represents the full population of the study. The cross-validation showed that the training set had 80% accuracy in predicting HES in the leak group. The misclassified spectra (20%) all had low concentrations of HES (below 0.66 g/mL), representing the prediction limit of the training set. The accuracy of the training set to predict the nonleak population was much higher, at 98%.

Leave a Reply

Your email address will not be published. Required fields are marked *