基于随机森林的台风风暴潮强度估计方法
Typhoon storm surge intensity estimation method based on Random Forest
侯琪 1李晓敏 2张杰 3张素铭 1赵堂麒 1杜雪雪1
作者信息
- 1. 中国石油大学(华东)海洋与空间信息学院,山东青岛 266580
- 2. 自然资源部第一海洋研究所,山东青岛 266061
- 3. 中国石油大学(华东)海洋与空间信息学院,山东青岛 266580;自然资源部第一海洋研究所,山东青岛 266061
- 折叠
摘要
为准确估计台风风暴潮强度,提出了一种基于随机森林(Random Forest,RF)的台风风暴潮强度估计方法.该方法采用斯皮尔曼(Spearman)相关性分析确定影响台风风暴潮强度的关键台风参数,将历史台风参数和台风风暴潮强度作为样本集,建立基于RF的台风风暴潮强度估计模型.利用我国沿海2000～2020年的55组历史台风参数和台风风暴潮强度数据集对所建立的台风风暴潮强度估计模型进行训练和测试.实验结果表明,RF算法估计的台风风暴潮强度精度均在80％以上,且高于C4.5算法,KNN算法和NBM算法估计的精度,验证了 RF算法估计台风风暴潮强度的可靠性.
Abstract
To accurately estimate the typhoon storm surge intensity,a typhoon storm surge intensity estima-tion method based on Random Forest(RF)is proposed.Spearman correlation analysis is used to determine the key typhoon parameters affecting the typhoon storm surge intensity.Taking the historical typhoon param-eters and typhoon storm surge intensity data as the sample set,the typhoon storm surge intensity estimation model based on RF is established.55 groups of historical typhoon parameters and typhoon storm surge in-tensity datasets in coastal China from 2000 to 2020 are used to train and test the established typhoon storm surge intensity estimation model.The results show that the accuracy of typhoon storm surge intensity esti-mated by RF algorithm is above 80％and higher than that estimated by C4.5,KNN and NBM algorithm,which verifies the reliability of RF algorithm for estimating typhoon storm surge intensity.
关键词
台风风暴潮强度/台风参数/随机森林/机器学习/Spearman相关性分析Key words
typhoon storm surge intensity/typhoon parameters/Random Forest/machine learning/Spearman correlation analysis引用本文复制引用
基金项目
国家重点研发计划项目(2022YFC3105102)
出版年
2024
NETL
NSTL
万方数据