摘 要：筛选出对病情发展有重要影响的因素，对新增确诊人数和新增死亡人数做出预测。通过随机森林的特征重要性筛选出对疫情发展影响最大的因素，使用 LSTM（Long Short Term Memory Network）建立预测模型。机场的繁忙程度对确诊人数影响最大，人口密度与死亡人数的关联性最大。美国地区的人口密度和机场交通情况对感染人数影响较大，从而影响年龄在80 岁以上老人的死亡率，但分析结果显示美国疫情发展已基本趋于稳中下降的态势。
关键词：COVID-19；影响因素；LSTM；感染数；死亡数
DOI:10.19850/j.cnki.2096-4706.2021.07.024
基金项目：国家自然科学基金（61863022)
中图分类号：R318；TP183 文献标识码：A 文章编号：2096-4706（2021）07-0091-04
Disease Prediction and Influencing Factors Analysis of COVID-19 Based on LSTM Model
HU Haiwen
（School of Mathematics and Physics，Lanzhou Jiaotong University，Lanzhou 730070，China）
Abstract：To screen out the factors that have an important impact on the development of the disease，and to predict the number of new confirmed cases and new deaths. Based on the importance of random forest characteristics，the most influential factors were screened out，and the LSTM（Long Short Term Memory Network）was used to establish the prediction model. The business of the airport has the greatest impact on the number of confirmed cases，and the population density has the greatest correlation with the number of deaths. The population density and airport traffic conditions in the United States have a great impact on the number of infected people，thus affecting the mortality of the elderly over 80 years old. However，the analysis results show that the development of the epidemic situation in the United States has basically tended to a steady decline.
Keywords：COVID-19；influencing factor；LSTM；number of infections；number of deaths
参考文献：
[1] TANG B，ZHANG X，LI Q，et al. The minimal COVID-19 vaccination coverage and efficacy to compensate for potential increase of transmission contacts，and increased transmission probability of the emerging strains [J/OL].BNE Public Health，2021：[2021-02-24]. https://www.researchsquare.com/article/rs-140717/v1.
[2] CDC. Covid data tracker [EB/OL].[2021-03-01]. https://covid. cdc.gov/covid-data-tracker/#datatracker-home.
[3] Johns Hopkins University & Medicine. COVID-19 Dashboard by the Center for Systems Science and Engineering （CSSE） at Johns Hopkins University （JHU） [EB/OL].[2021-03-01].https://coronavirus. jhu.edu/map.html.
[4] RANDHAWA G，SOLTYSIAK M，EL ROZ H，et al. Machine learning using intrinsic genomic signatures for rapid classification of novel pathogens：COVID-19 case study [J/OL].PLOS One，2020，15（4）：[2021-03-01].https://journals.plos.org/plosone/ article?id=10.1371/journal.pone.0232391.
[5] 胡小亮，路方红，刘振东 , 等 . 原发性高血压患者血压晨 峰与心血管疾病危险因素相关性分析 [J]. 中国循环杂志，2012，27 （2）：95-98.
[6] 汤正义，李向利，张炜，等 .2 型糖尿病神经病变的发病 情况与有关因素相关性分析 [J]. 中华内科杂志，2003（12）：47- 48.
[7] 黄梅，朱焱 . 基于随机森林特征重要性的 K- 匿名特征优 选 [J]. 计算机应用与软件，2020，37（3）：266-270.
[8] 李俊卿，李秋佳，石天宇，等 . 基于随机森林重要性的 LSTM 网络风电功率缺失数据补齐 [J]. 电器与能效管理技术，2018 （13）：47-52+58.
[9] 欧阳红兵，黄亢，闫洪举 . 基于 LSTM 神经网络的金融时 间序列预测 [J]. 中国管理科学，2020，28（4）：27-35.
[10] ALIMADADI A，ARYAL S，et al. Artificial intelligence and machine learning to fight COVID-19 [J].Physiological Genomics， 2020，52（4）：200-202.
[11] WANG P P，ZHENG X Q，LI J Y，et al. Prediction of epidemic trends in COVID-19 with logistic model and machine learning technics [J/OL].Chaos，Solitons & Fractals，2020，139：[2021-03-01]. https://doi.org/10.1016/ j.chaos.2020.110058.
作者简介：胡海文（1995—），女，汉族，甘肃嘉峪关人，硕 士研究生在读，研究方向：应用统计。