Machine learning approach to an otoneurological classification problem.

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Oto-neurological Disorders

Background:

• Accurate classification of otoneurological diseases is crucial for effective patient management.
• Existing diagnostic methods may benefit from advanced computational approaches.
• Machine learning offers potential for improving diagnostic accuracy in complex medical fields.

Purpose of the Study:

• To evaluate the efficacy of various classification methods for otoneurological disease classification.
• To investigate the performance of the Half-Against-Half (HAH) architecture in this diagnostic context.
• To identify the most accurate classification approach for the given otoneurological dataset.

Main Methods:

• Application of thirteen distinct classification algorithms, including Support Vector Machines (SVM), k-Nearest Neighbour (k-NN), Naïve Bayes (NB), and Multinomial Logistic Regression (MNLR).
• Implementation of the Half-Against-Half (HAH) architecture in conjunction with SVM, k-NN, and NB methods.
• Comparative analysis of classification accuracies achieved by each method on the otoneurological disease dataset.

Main Results:

• The Half-Against-Half Support Vector Machine (HAH-SVM) with a linear kernel yielded the highest classification accuracy at 76.9%.
• Other methods, including HAH-k-NN, HAH-NB, and Multinomial Logistic Regression (MNLR), achieved accuracies exceeding 60%.
• The obtained 77% accuracy represents a significant and competitive result compared to prior research using the same dataset.

Conclusions:

• The HAH-SVM approach demonstrates superior performance for otoneurological disease classification within this study.
• The HAH architecture shows promise as an effective framework for medical diagnostic tasks.
• Further research validating these findings on larger and diverse datasets is warranted.