Skip to main content
. 2018 Oct 15;8:15229. doi: 10.1038/s41598-018-33449-0

Figure 5.

Figure 5

Characterization of a Supported-Vector-Machine (SVM) prediction model for the identification of CLN2 based on 8 CSF-biomarkers. (a) Receiver-operating-characteristic (ROC)-analysis shows the sensitivity and specificity of the model with an area-under-the-curve (AUC) of 0.92 and (a) confidence interval (CI) of 0.667–1. (b) SVM-predicted class probabilities. A value of >0.5 leads to a CLN2 disease classification, a value of <0.5 to a classification as a control. (c) Results of the predictive classification of the SVM-model for the test and validation cohort. B12 and N4 are misclassified in the test cohort, B21 is misclassified in the validation cohort.