Skip to main content
. Author manuscript; available in PMC: 2022 Jan 1.
Published in final edited form as: Stroke. 2020 Dec 10;52(1):181–189. doi: 10.1161/STROKEAHA.120.030663

Figure 4.

Figure 4.

Random forest importance scores identifying important variables for predicting TOAST cardioembolic stroke.

The best performing model was random forest, which showed that atrial fibrillation and age were the most important variables for discriminating cardioembolic stroke. Next most important were congestive heart failure, hypokinetic left ventricular segment, mitral annulus calcification, gender, and dilated cardiomyopathy.