Machine Learning Recognition of Otoneurological Patients by Means of the Results of Vestibulo-Ocular Signal Analysis

We distinguished a group of otoneurological patients from healthy subjects on the basis of machine learning methods applied to signal analysis results calculated in our earlier research. We classified them to investigate, which methods are the most efficient to separate the two classes from each other. Decision trees and support vector machines yielded the highest average accuracies of 89.8% and 89.4% being 1-5% better than others.