Precision Medicine Towards Women’s Multiple Sclerosis Progression Using Machine Learning Models

Over 2.8 million people worldwide are affected by multiple sclerosis, an autoimmune disease in which the immune system attacks the central nervous system. Most individuals with MS experience some degree of neurological disability, and women have an approximately fourfold risk of developing MS compared to men. Diagnosing MS and identifying its subtypes remain challenging and time-consuming, due to varying clinical evaluations across patients. This study investigates whether ensemble optimization improves multi-class MS subtype classification accuracy compared to individual ML models, addressing the research gap in existing approaches that lack ensemble optimization. We preprocessed a publicly accessible, fully anonymized dataset consisting of 273 patient health records to ensure completeness and balance across classes. ML algorithms representing tree-based, neural, and instance-based learning paradigms were developed for multi-class classification in disability score; prevented from overfitting using dropout and regularization; and evaluated using precision, recall, and F1-score. To combine predictions, we implemented ensemble averaging using performance-based weighted voting. The results indicate that assigning higher weights to better-performing models resulted in an accuracy improvement of approximately 2% over simple averaging. Among individual models, the random forest achieved the highest classification accuracy of 90% and the most reliable F1-score. While larger datasets are needed to meet the clinical requirement, these findings demonstrated the potential of machine learning to support clinical decision-making in MS diagnosis, which may be particularly impactful given the higher prevalence of MS in women. The study also highlighted the importance of incorporating gender- and lifestyle-related factors into future clinical and computational studies.