| 北京地区蚊虫密度变化气象预测方法研究 |
| |
| 引用本文： | 姜江,叶彩华,刘美德,尤焕苓,乔媛,夏江江,佟颖,张勇,阎婷,李秋红,刘婷,周小洁,曾晓芃.北京地区蚊虫密度变化气象预测方法研究[J].气象科技,2022,50(4):584-593. |
| |
| 作者姓名： | 姜江 叶彩华 刘美德 尤焕苓 乔媛 夏江江 佟颖 张勇 阎婷 李秋红 刘婷 周小洁 曾晓芃 |
| |
| 作者单位： | 北京市气象服务中心，北京100089;北京市疾病预防控制中心，北京 100013;中国科学院大气物理研究所，北京 100029 |
| |
| 基金项目： | 首都卫生发展科研专项（No.2018 2 3015）、北京市预防医学研究中心科研培育项目（No.2016 BJYJ 08）资助 |
| |
| 摘 要： | 本文利用蚊虫密度监测数据及气象资料，分析了2009—2019年北京市及其14个区的蚊虫密度与气象条件间的关系，并基于多元回归、支持向量机和随机森林3种经典的机器学习回归方法进行了蚊虫密度预测。结果表明：北京地区蚊虫密度呈周期性的波动，各区多年平均值在0.35~2.54只/(灯·h)之间，高峰值集中出现在7月中旬到8月中旬，与北京地区气温最高和降水最集中的时期非常吻合。采用机器学习方法，尝试了4种输入因子方案，并利用均方根误差和平均绝对百分误差两种方法进行预测效果检验，显示蚊虫数据相对较稳定的地区，如平谷、门头沟、大兴、海淀等地，预测效果相对更优。在3种方法中，支持向量机方法对2019年5月下旬的预测效果非常好，而多元回归与随机森林的预测效果则在2019年5—10月整体上表现得更为稳定。
|
| 关 键 词： | 北京 蚊虫密度 机器学习预测 |
| 收稿时间： | 2021/9/29 0:00:00 |
| 修稿时间： | 2022/2/23 0:00:00 |
Study on Meteorological Prediction Method of Mosquito Density in Beijing |
| |
| JIANG Jiang,YE Caihu,LIU Meide,YOU Huanling,QIAO Yuan,XIA Jiangjiang,TONG Ying,ZHANG Yong,YAN Ting,LI Qiuhong,LIU Ting,ZHOU Xiaojie,ZENG Xiaofan.Study on Meteorological Prediction Method of Mosquito Density in Beijing[J].Meteorological Science and Technology,2022,50(4):584-593. |
| |
| Authors: | JIANG Jiang YE Caihu LIU Meide YOU Huanling QIAO Yuan XIA Jiangjiang TONG Ying ZHANG Yong YAN Ting LI Qiuhong LIU Ting ZHOU Xiaojie ZENG Xiaofan |
| |
| Abstract: | This paper analyzes the relationship between mosquito density and meteorological conditions from 2009 to 2019 in Beijing and its 14 districts based on three machine learning methods. The result shows that the mosquito density fluctuates periodically from May to October each year. The average is between 0.35 to 2.54 per lamp·hour, and the peak appears in mid July to mid August, corresponding to the period of highest temperature and most precipitation in Beijing. We choose Multiple Regression, Support Vector Machine and Random Forest to predict the mosquito density for ten days with different input factors. RMSE and MAPE are used to test the prediction effect. It turns out that it is relatively better in areas where the mosquito density is stable, such as Pinggu, Mentougou, Daxing, Haidian and so on. In addition, among the three methods, the support Vector Machine Method has a very good prediction effect in late May 2019, while the prediction effect of Multiple Regression and the Random Forest is more stable from May to October 2019. |
| |
| Keywords: | Beijing mosquito density machine learning prediction |
|
| 点击此处可从《气象科技》浏览原始摘要信息 |
| 点击此处可从《气象科技》下载免费的PDF全文 |
|
