Cómo citar
Resumen
In Japan, 2,000 to 3,000 drowning accidents occur every summer season at major bathing beaches. In order to prevent drowning accidents, beachgoers themselves need to be aware of the dangers and avoid them. As a way to do this, bathing beaches provide daily risk levels regarding swimming conditions to beachgoers using three levels of beach safety flags. However, the risk levels are determined subjectively and empirically by lifesavers and beach administrators based on weather and sea conditions. The characteristics of past drowning accidents are not taken into account. In this study, we suggest an objective method of determining the risk levels based on the probability of drowning accidents. We have created an AI model that can predict the probability of drowning accidents with high accuracy using a total of 53 features such as usage, weather and sea conditions of a study beach in Japan. This method enables appropriate judgment of swimming conditions to prevent many drowning accidents. The reliability of the model was examined using XAI, and it was found that time series of rescue factors were important in predictions. On the other hand, the accuracy decreased when the created AI model was applied to other beaches. It was thought to be caused by differences in the natural environment such as waves and wind.Referencias
Akobeng AK. 2007. Understanding diagnostic tests 3: Receiver operating characteristic curves, Acta
Paediatr, 96(5), 644-647.
Arrieta, A. B., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Barbado, A., Garcia, S., Gil-
Lopez, S., Mo-lina, D., Benjamins, R. and Herrera, F. 2020. Explainable Artificial Intelligence
(XAI): Concepts, taxonomies, opportunities, and challenges toward responsible AI. Information
fusion, 58, 82-115.
Ishikawa, T., Komine, T., Aoki, S. and Okabe, T. 2014. Characteristics of Rip Current Drowning on the
Shores of Japan, Journal of Coastal Research, 72, 44-49.
Komine, T., Tanaka, H., Takyu, H., Kinoshi, T., Gotoh, S., Sone, E., Sagisaka, R., Ishikawa, T. and
Shimazaki, S. 2015. Effectiveness of surf lifesaver on OHCA occurred by drowning on the beaches
in Japan, 8th Asian Conference for Emergency Medicine.
Lundberg, S. M. and Lee, S. I. 2017. A unified approach to inter-preting model predictions, Advances in
neural infor-mation processing systems, 30.
Lundberg, S. M., Erion, G. G. and Lee, S. I. 2018. Consistent indi-vidualized feature attribution for tree
ensembles. arXiv preprint arXiv:1802.03888.
N. V., Bowyer, K. W., Hall, L. O., and Kegelmeyer, W. P. 2002. SMOTE: synthetic minority oversampling
technique, Journal of artificial intelligence research, 16, 321-357.
Q. M. Guolin Ke, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye and Te-Yan Liu.
LightGBM: A Highly Efficient Gradient Boosting Decision Tree, Advances in Neural
Information Processing Systems, 30.
Shimada, R., Ishikawa, T., Komine, T. 2020. Outbreak Factors of Rip Current Accident at Onjuku Breach
in Japan, Japan Society of Civil Engineers (Ocean engineering), Vol.75, pp. I_695-I_700.
Snoek, J., Larochelle, H., & Adams, R. P. 2012. Practical bayesian optimization of machine learning
algorithms, Advances in neural information processing systems, 25.
W. J. Youden. 1950. Index for rating diagnostic tests, Cancer, 3(1), 32–35.
Esta obra está bajo una licencia internacional Creative Commons Atribución 4.0.
Derechos de autor 2023 Haruki Toguchi, Ryo Shimada, Toshinori Ishikawa, Tsutomu Komine