<![CDATA[The accurate classification of migraine subtypes is a complex challenge in neurology, hindered by symptomatic similarities between types. This complexity necessitates advanced computational tools to support diagnostic precision. This study aims to develop and evaluate an optimized soft voting ensemble classifier to automate this multi-class classification task effectively. The methodology involved training eight base models—including Neural Network, Random Forest, and Gradient Boosting—on a publicly available migraine dataset, with an 80-20 train-test split. The top three performers were integrated into a soft voting ensemble, which aggregates their predicted probabilities to enhance decision robustness. Model performance was rigorously assessed using accuracy, precision, recall, F1-score, and AUC-ROC metrics. The results demonstrated that the proposed ensemble achieved superior performance, with an accuracy of 91.67% and an F1-score of 91.50%, outperforming all constituent models. Furthermore, the ensemble attained near-perfect AUC-ROC values across multiple classes, confirming its strong discriminatory capability. The study concludes that the soft voting ensemble is a highly effective and reliable approach for migraine subtype classification, offering significant potential as a decision-support tool in clinical environments. Future work will focus on hyperparameter optimization, explainability, and validation with larger multi-centric datasets to facilitate clinical adoption.]]>
Copyrights © 2025
