| 基于随机森林的复坡堤越浪量预测研究 |
| |
| 引用本文： | 胡原野,王收军,陈松贵,柳叶,王家伟,田昀艳.基于随机森林的复坡堤越浪量预测研究[J].海洋学报,2021,43(10):106-114. |
| |
| 作者姓名： | 胡原野 王收军 陈松贵 柳叶 王家伟 田昀艳 |
| |
| 作者单位： | 1.天津理工大学 机电工程国家级实验教学示范中心，天津 300384 |
| |
| 基金项目： | 国家自然科学基金（52001149，52039005，51861165102）；中国科协青年人才托举工程（2018QNRC001）；中央级公益性科研院所基本科研业务费（TKS20200204，TKS20210102，TKS20210110）；天津市科技计划项目 （17PTYPHZ00080） |
| |
| 摘 要： | 针对复坡堤越浪量的计算问题,提出了采用随机森林算法预测越浪量的方法。首先,通过对欧洲CLASH数据集进行筛选,挑选出符合复坡堤越浪量预测的数据;其次,对数据做无量纲化处理,建立以随机森林为基础的复坡堤越浪量预测模型,并通过网格搜索(GridSearchCV)方法对模型进行调参以改善模型的性能;最后,利用决定系数R~2来评估模型的精度,并将随机森林模型与集成神经网络模型做了预测能力的对比,同时还给出了随机森林模型各个特征参数对预测精度的重要性。结果显示,随机森林模型的决定系数为92.7%,集成神经网络模型的决定系数为87.7%,表明随机森林模型对越浪量具有更强的学习和预测能力。通过对特征重要性的分析,墙顶高程对模型预测精度的影响最大,堤顶高程次之,堤脚宽度影响最小。
|
| 关 键 词： | 随机森林 越浪量 复坡堤 决定系数 特征重要性 预测 |
| 收稿时间： | 2020-06-27 |
Overtopping prediction for composite slope breakwater based on random forest method |
| |
| Institution: | 1.National Demonstration Center for Experimental Mechanical and Electrical Engineering Education, Tianjin University of Technology, Tianjin 300384, China2.National Engineering Laboratory for Port Hydraulic Construction Technology, Tianjin Research Institute for Transport Engineering, Tianjin 300456, China |
| |
| Abstract: | Aiming at the problem of calculating overtopping of the composite slope breakwater, a prediction model of the overtopping for the composite slope based on the random forest method is proposed. Firstly, by filtering the European CLASH data set, the data consistent with the prediction of overtopping of the composite slope breakwater are selected. Secondly, after dimensionless processing of the data, overtopping prediction model is established based on random forest method, and improved by adjusting the model parameters according to GridSearchCV. Finally, the coefficient of determination R2 is used to evaluate the accuracy of the model, and the prediction ability of the model is compared with the ensemble neural network model. The effect of each feature parameter of the random forest model on the prediction accuracy is assessed. The results show that the coefficient of determination of the random forest model is 92.7%, and the coefficient of determination of the ensemble neural network model is 87.7%, indicating the random forest model has a stronger prediction ability for predicting overtopping. Wall height with respect to static water level has the greatest influence on the prediction accuracy of the model, the height of the top of the embankment is the second, and the width of the foot of the embankment least. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《海洋学报》浏览原始摘要信息 |
| 点击此处可从《海洋学报》下载免费的PDF全文 |
|
