Predictive Modeling for Heatstroke Risk Forecasting Integrating Physiological Features Using Ensemble Classifier

Heatstroke is life-threatening, with rising mortality rates attributed to increasing global temperatures. Acute discernment of heatstroke symptoms remains imperative for effective interventions. We proposed an advanced postprocessing technique to forecast heatstroke risk, enhancing machine learning algorithms based on physiological features. We implemented a cuttingedge ensemble model that amalgamates the statistical and forecasting models’ outputs. Utilizing statistical models such as Random Forest, Ada-Boost, and Support Vector Machine, we attained an accuracy of 96.19±1.6%. Employing forecasting models like Auto-Regressive Integrated Moving Average (ARIMA), Seasonal ARIMA, and Prophet yielded 96.69±1.45% classification accuracy. The ensemble of statistical and forecasting models exhibited outstanding metrics, including precision of 99.16±0.15%, recall of 98.93±0.14%, and F1-score of 98.68±0.10% in thermal comfort forecasting. This groundbreaking endeavor holds promise for advancing the state-of-the-art in heatstroke prevention, augmenting the accuracy and reliability of the predictive models, and consequently, benefiting public health initiatives.

This is an Accepted Manuscript of a book chapter published by CRC Press in Activity, Behavior, and Healthcare Computing on [In Press], available online: https://www.routledge.com/9781032639185