Confounding Values in Decision Trees Constructed for Six Otoneurological Diseases

In this study, we examined the effect of example cases with confounding values on decision trees constructed for six otoneurological diseases involving vertigo. The six diseases were benign positional vertigo, Menière’s disease, sudden deafness, traumatic vertigo, vestibular neuritis, and vestibular schwannoma. Patient cases with confounding values were inserted into original vertigo data and decision trees were constructed. Confounding values made classification tasks more difficult and decreased true positive rates and accuracies of decision trees. Despite decreased true positive rates and accuracies, new decision trees organised confounding values in a reasonable way into the reasoning process. The occurrence of confounding values simulates better the real life classification tasks.