Application of Machine Learning to the Prediction of WBGT

In Japan, the number of emergency patients and deaths from heatstroke has been increasing due to abnormal climate as global warming. For preventing heat illness prevention, the Japanese Ministry of the Environment has specified Wet Bulb Globe Temperature (WBGT) as a heat illness risk on the website since 2006. WBGT represents a heat stress index based on the important three factors; temperature, humidity and radiation which effects human heat balance. Depending on environmental situation and condition, WBGT index is calculated from dry-bulb temperature, wet-bulb temperature and globe temperature. In this research, incorporating cycle and autocorrelation of WBGT and using machine learning, we try to build WBGT prediction model which does not take meteorological information as an input variable. Based on the predicted WBGT, we assess the heat risk ranks: danger, severe warning, warning, caution and almost safe. The proposed method will be applied for predicting WBGT of Osaka, and through the comparison with the conventional method, which are adopted by the Japanese Ministry of the Environment, the effectiveness of the proposed model is investigated in terms of prediction ability of WBGT index and heat risk assessment using WBGT.