Date of Award
Fall 2023
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Electrical and Computer Engineering
First Advisor
Povinelli, Richard
Second Advisor
Ababei, Cristinel
Third Advisor
Frigo, Frederick
Abstract
Benign Paroxysmal Positional Vertigo (BPPV) is one of the most common causes of dizziness. Especially for people over 45, the risk of BPPV is substantial. On the other hand, BPPV is often misdiagnosed and may require expensive examinations. This thesis introduces a prediction model based on machine learning to quickly, inexpensively, and accurately diagnose BPPV. The thesis starts by introducing BPPV and the statistics of BPPV misdiagnosis. Then, a patient survey is introduced. The patient survey includes 50 BPPV-related questions, which are used as training data for the machine learning model. Logistic Regression, Decision Tree, and Naïve Bayes were compared for machine learning models and their results were discussed. Three machine learning approaches are explored, logistic regression, decision tree, and naïve Bayes with cross validation accuracies of 89.8%, 81.9%, and 75.1%, respectively.