This article focuses on the design of typhoon risk assessment and pre-alarm system under the condition of El Nino. In view of the remarkable influence of El Nino phenomenon on typhoon activities and the challenges faced by traditional assessment and pre-alarm methods, this article makes an in-depth study with the help of Machine Learning (ML) technology. By collecting El Nino related data, typhoon data and auxiliary data, after data preprocessing, the assessment index system is determined from meteorological, geographical and socio-economic levels. In this article, ML algorithm such as neural network is selected to build a risk assessment model, and a typhoon risk pre-alarm system including data layer, model layer and application layer is designed and implemented. The experimental results show that the prediction accuracy of the model is high. Taking the confusion matrix as an example, the correct proportion of high-risk grade prediction is 79%. The accuracy of pre-alarm increases with the increase of pre-alarm time, and the accuracy of pre-alarm 60 hours in advance can reach 90%. Practical application cases show that the system can effectively control casualties and economic losses. For example, in the case of typhoon "Jongdari" in Binhai City in 2020, the number of seriously injured people decreased by about 87.5% and the economic losses decreased by about 80%, which fully verified the effectiveness of the system.
Yang et al. (Sun,) studied this question.