Tweets and AI models Unveiling HittbooT risks in urban areas

Heat card is a significant health risk, especially during extreme temperature conditions. As global temperatures rise as a result of climate change, the frequency and severity of heat waves have increased, so that vulnerable populations have brought a greater risk.

This shift underlines the need for effective, real -time methods for early detection and response to risks in the field of heat material, which guarantees timely intervention and reduced impact of these rising threats.

Although earlier studies have emphasized the potential of reports on social media, such as tweets, to offer real -time insights into various events, their application was not investigated when detecting heat material risks.

To tackle this gap, a team of researchers, led by professor Sumiko Anno of the Graduate School of Global Environmental Studies, Sophia University, Japan, together with Dr. Yoshitsugu Kimura, Yanagi Pearls, Japan, and Dr. Satoru Sugita, Chubu University, Japan has used the potential to combine posts on social media and transformers -based learning models to detect heat material risks in Nagoya City, Japan.

Their findings were published in Scientific reports.

The researchers used in-depth learning models based on transformers, including Bert, Roberta and Luke Japanese Base Lite, together with a machine learning model (support vector machine or SVM) to identify tweets that contain the word “hot” in Japanese.

The team successfully collected around 27,040 tweets for a period of five years with the help of the Twitter API. By processing the text data in advance and applying advanced techniques for deep and machine learning, the models were trained and refined to identify tweets related to heat material events. These models were evaluated with the help of important performance statistics such as accuracy, precision, recall and F1 score.

Among the tested models, Luke Japanese base Lite reached the highest performance statistics with an accuracy of 85.52%, followed by Bert-Base (84.04%) and Roberta-Base (83.88%). While the SVM basis line model showed the lowest performance, with an accuracy of 72.73%.

In addition, the use of visualizations and animated video from Time-Space showed the potential for real-time events-based surveillance. By mapping the locations of heat-related medical emergency evacuations and matching them with geotagged tweets, the study showed how social media data can offer an early warning system for risks in the field of heat stroke in urban environments.

Prof. Anno explains: “By using messages on social media, we can improve the supervisory systems for public health and the early detection of risks in the field of heating.”

The research emphasizes the potential of combining Japanese tweets and transformer -based propraned language models for public health supervision. Luke’s superior performance When detecting heat stroke-related tweets, their viability suggests when monitoring heat material risks during heat waves.

In addition, the visualisations of the time-space showed how social media can be integrated with emergency aid data to serve as an effective tool for early detection for extreme weather conditions.

This research opens the door to future applications of in-depth learning and social media messages for real-time health monitoring systems. As climate change increases, the ability of early detection and response to risks in the field of heat stroke can become a crucial tool in protecting public health.

Looking ahead, the team is planning to set up an early warning system for a heat stroke in the Aichi prefecture, with the aim of ultimately extending this system to a national warning system for Japan. Important steps in achieving this are collaboration with local authorities to collect heat stroke data and to perform spatiotemporal analyzes in all prefectures.

“Our methodology can be expanded and adapted for monitoring emerging and re -emerging infectious diseases, which broadens its application in public health monitoring,” concludes Prof. dr. Anno.