Our objective was to predict the number of heat stroke patients from meteorological and Twitter data.
Materials and Methods: Using 2015-2017 data for Kanagawa and Osaka, we fitted a formula to predict the number of heat stroke patients from daily mean temperature and the number of tweets (Twitter data points) about heat stroke. Then, to verify the accuracy of the formula, we applied it to 2016-2017 data for Tokyo and 2015-2017 for Kanagawa and Osaka.
Results: The coefficient of correlation between the actual and predicted numbers of heat stroke patients in Tokyo was 0.9726, and prediction accuracy was improved by including number of tweets, even when applying the formula to another area.
Discussion and Conclusion: The number of heat stroke patients can be predicted in real time from daily mean temperature and Twitter data, which could allow protective measures to be taken.
View full abstract