Identifying the neurodevelopmental and psychiatric signatures of genomic disorders associated with intellectual disability: a machine learning approach

Table 3 Final model performance on held-out test dataset

Model	Brier score	AUC ROC	AUC ROC difference	Probability of direction
Random forest	0.188 [0.121, 0.243]	0.915 [0.838, 0.98]	–	–
Penalised LR	0.183 [0.121, 0.251]	0.904 [0.82, 0.981]	0.011 [− 0.099, 0.122]	0.843
ANN	0.186 [0.152, 0.225]	0.883 [0.787, 0.963]	0.031 [− 0.087, 0.151]	0.619
RBF SVM	0.21 [0.137, 0.284]	0.897 [0.814, 0.968]	0.018 [− 0.089, 0.124]	0.757

Values shown are bootstrapped performance and the 95% confidence interval of the measure (AUROC and Brier Score), and difference in AUROC between the random forest and other ML models, with its 95% confidence interval, and the probability of direction for the AUROC difference

ISSN: 2040-2392