Prediction of robotic neurorehabilitation functional ambulatory outcome in patients with neurological disorders

Table 3 Performance for predicting effectiveness of RAGT using (a) Clinical data and all sessions for improvement or not by ten-fold cross-validation, and (b) Using clinical data and all sessions to predict three levels of FAC change by ten-fold cross-validation

Algorithms	Accuracy	Sensitivity	Specificity	AUC
(a) Prediction performance of model 2 using clinical data and all sessions for improvement or not by ten-fold cross-validation
Random Forest	0.879 ± 0.079	1.000 ± 0.000	0.767 ± 0.161	0.981 ± 0.043
Logistic Regression	0.853 ± 0.101	0.933 ± 0.141	0.767 ± 0.141	0.918 ± 0.061
Support Vector Machine	0.836 ± 0.124	0.855 ± 0.166	0.850 ± 0.123	0.919 ± 0.120
Decision Tree	0.618 ± 0.087	0.626 ± 0.208	0.717 ± 0.284	0.695 ± 0.160
XGBoost	0.854 ± 0.072	0.933 ± 0.141	0.817 ± 0.166	0.937 ± 0.070
(b) Prediction performance of model 3 using clinical data and all sessions to predict three levels of FAC change by ten-fold cross-validation
Random Forest	0.747 ± 0.074	0.567 ± 0.077	0.846 ± 0.049	0.814 ± 0.093
Logistic Regression	0.535 ± 0.114	0.441 ± 0.147	0.732 ± 0.065	0.675 ± 0.127
Support Vector Machine	0.522 ± 0.132	0.391 ± 0.106	0.720 ± 0.084	0.724 ± 0.131
Decision Tree	0.499 ± 0.065	0.329 ± 0.043	0.667 ± 0.029	0.485 ± 0.038
XGBoost	0.731 ± 0.079	0.580 ± 0.116	0.842 ± 0.051	0.768 ± 0.119

ISSN: 1743-0003