A Predictive Machine Learning Analysis for Stroke Prediction

Stroke impact on an individual's health remains alarming around the globe, and requires analysis and treatment on time. This paper discusses the design and implementation of IoT based health device incorporating highly sophisticated machine learning techniques for realtime stroke prediction. The system is comprised of temperature, pulse, and ECG sensors connected through an ESP32 microcontroller. The data is collected and stored on the ThingSpeak cloud for further analysis. The ANN algorithm within MATLAB showcased the highest stroke prediction accuracy at 92%, demonstrating strong performance for identifying at-risk patients. For comparison, the same dataset was utilized to train a Generative AI (Gen AI) model and results showed an 88% accuracy, which is lower than ANN due to structured health data complexity. Beyond showing the promise of Gen AI in healthcare analytics, this demonstrates how more complex models can fail to outperform simpler models for certain applications. The device provides real time alarms and predictive information, which is a step towards more advanced healthcare systems.