Migraine Categorization using the Scatter Search and Random Forest Classifier

Migraine attacks manifest through recurring episodes of intense headaches accompanied by autonomic nervous system dysfunction indicators, leading to various symptoms. Automatic classification of various classes of migraine is very much essential for the secondary opinion. This study categorizes a migraine dataset into seven distinct classes using features obtained through scatter search. To address the class balance issue, Synthetic Minority Oversampling Technique (SMOTE) was applied. Further, the optimal features were selected using Scatter Search. Subsequently, classification was executed using the Random Forest (RF) classifier. The proposed model was tested on a migraine dataset containing 400 instances and 24 features, resulting in a feature reduction of around 50%. Remarkably, the achieved accuracy of 98.26% surpassed that of the raw dataset. As a result, the proposed model outperforms existing methodologies, potentially offering an additional perspective for medical professionals.