| 基于遗传算法的特征选择方法在短时强降水预报中的应用北大核心 |
| |
| 引用本文： | 耿焕同,戴中斌,沈阳.基于遗传算法的特征选择方法在短时强降水预报中的应用北大核心[J].气象科学,2023,43(1):126-134. |
| |
| 作者姓名： | 耿焕同 戴中斌 沈阳 |
| |
| 作者单位： | 南京信息工程大学 计算机与软件学院, 南京 210044;江苏省气象台 南京大气科学联合研究中心, 南京 210019 |
| |
| 基金项目： | 北极阁开放研究基金—南京大气科学联合研究中心(NJCAR2018MS05);国家自然科学基金资助项目(51977100) |
| |
| 摘 要： | 利用江苏省13个气象观测站历史上短时强降水观测资料,用遗传算法进行特征选择,选定影响短时强降水的950 hPa假相当位温、700 hPa比湿、500 hPa比湿、对流有效势能(Convective Available Potential Energy,CAPE)等14个特征为主要因素,将是否为短时强降水抽象成二元分类问题。借助机器学习中CART决策树算法进行分类分析,构建便于使用的短时强降水预报规则集。实验部分,随机选择5816条样本进行训练模型,得到适合江苏地区的短时强降水规则集,利用剩余的1454条数据进行实际检验,模型的短时强降水预报准确率为91.35%,非强降水预报准确率为97.11%,较特征选择之前分别提升了8.66%和1.05%。
|
| 关 键 词： | 短时强降水 预报模型 特征选择 遗传算法 CART决策树 |
| 收稿时间： | 2020/10/27 0:00:00 |
| 修稿时间： | 2021/3/1 0:00:00 |
Application of feature selection method based on genetic algorithm in short-term heavy precipitation forecast |
| |
| GENG Huantong,DAI Zhongbin,SHEN Yang.Application of feature selection method based on genetic algorithm in short-term heavy precipitation forecast[J].Scientia Meteorologica Sinica,2023,43(1):126-134. |
| |
| Authors: | GENG Huantong DAI Zhongbin SHEN Yang |
| |
| Abstract: | Based on the historical observation data of short-term heavy precipitation at 13 meteorological stations in Jiangsu Province, the genetic algorithm was used for feature selection, and 14 features, such as pseudo equivalent potential temperature at 950 hPa, specific humidity at 700 hPa, specific humidity at 500 hPa and Convective Available Potential Energy(CAPE), were selected as the main factors to abstract the short-term heavy precipitation into a binary classification problem. With the help of CART algorithm in machine learning, a set of short time heavy precipitation forecast rules for the use of forecasters is constructed. In the experiment part, 5 816 samples were randomly selected to train the model, and the rule set of short-time heavy rainfall suitable for Jiangsu was obtained. The remaining 1 454 data were used for practical test. The prediction accuracy of short-time heavy precipitation was 91.35% and that of non-heavy precipitation was 97.11%, which are 8.66% and 1.05% higher than that before feature selection. |
| |
| Keywords: | short-term heavy precipitation forecasting model feature selection genetic algorithm CART decision tree |
| 本文献已被 维普 等数据库收录！ |
| 点击此处可从《气象科学》浏览原始摘要信息 |
| 点击此处可从《气象科学》下载免费的PDF全文 |
|
