Data-driven analysis of stroke-related factors and diagnostic prediction

The exponential growth of the Internet of Things (IoT), Big Data, and Artificial Intelligence (AI) technologies has facilitated the acquisition of medical data, thereby paving the way for advanced analysis and diagnosis in healthcare. This has gained significance, particularly in the context of stroke, the incidence of which has been skewing younger due to lifestyle changes, posing a significant threat to public health. This paper undertakes a data-driven exploration of the contributing factors to strokes, encompassing data collection, preprocessing (managing missing values), and a detailed correlation analysis involving visualization and correlation matrices. Existing stroke diagnostic techniques predominantly hinge on medical expertise, a resource that may be scant in certain regions, resulting in delayed diagnosis and disease progression. To address this issue, this study investigates the potential of machine learning methods for stroke diagnosis. The paper employs logistic regression, support vector machines, and random forests for model training and testing, which involves data standardization, partitioning into training and testing sets, model construction, and evaluation. The results indicate that all three models demonstrate an accuracy rate exceeding 90%, underscoring the viability of these methods for stroke diagnosis.