摘要
为了更合理地评估物流系统引起疫情传播的风险,提供了一种基于随机森林回归、K近邻(K-Nearest Neighbor,KNN)及专家判断法的风险组合评估算法。通过构建传播疫情风险的组合评估基础方法模型,对目标地区系统风险等级进行预判,对人员、交通工具、周转工具、货物包装、产品及系统风险等级进行仿真。首先构建随机森林回归模型预测感染者数据,构建KNN聚类及专家打分预判等级的风险组合算法测度风险等级。之后利用留出法按8∶2将7国省级数据分为训练集和预测集来训练模型,得到最优随机森林回归模型;利用专家打分法设定低、中、高风险区间,按照此区间预设1000条数据,且随机打乱其中的5000个数值构成矩阵,作为KNN的假设数据集,训练得到最优KNN模型,准确率(Accuracy,ACC)为0.975。最后对非精准国家吉尔吉斯斯坦共和国进行实证仿真,得到五要素的风险值,其物流系统整体风险等级为高风险。该组合算法是物流系统疫情防控的技术突破点,可为决策提供数据支撑。
This paper explores the methods to prevent the epidemic spread in the logistics system processes,quickly and quantitatively analyzes the risk’s degrees in advance,and makes decisions for prevention and control of the epidemic spread.Through simulating we predicted the system risk level of the target area,and the risk levels of personnel,vehicles,turnover tools,goods packaging,and products by constructing the basic method model of combined assessment of epidemic risk.Firstly,we constructed a risk combination algorithm based on random forest,K-Nearest Neighbor(KNN),and an expert scoring method.The random forest algorithm predicted the number of infected people in national provinces or regions,the expert scoring method predicted the risk level of epidemic spread,and the expert scoring method and KNN combination algorithm measured the risk level.Then,based on the data from the provinces of the seven countries,we obtained the optimal parameter model of random forest regression by randomly dividing the data into a training set and prediction set according to 8:2 with the reserve method of 15 degrees of freedom.The values of[0,0.2],(0.2,0.8],and(0.8,1]occupied 80%,16%,and 4%respectively according to the interval of low risk[0,0.3),medium risk[0.3,0.9),high risk[0.9,1].Based on the matrix which was composed of 5000 values through the preset 1000 pieces of data randomly formed into chaos with the labeling L=2,M=1,and H=0 as KNN’s hypothetical historical data set,we obtained the optimal parameter combination by training the KNN model.Finally,the results show that the risk level of its logistics system,five elements and among three of them in the imprecise Kyrgyzstan Republic was high degree by the empirical simulation.We can conclude that the combined algorithm in this study is a technical breakthrough in epidemic prevention and control of logistics systems.It can provide data support for decision-making in advance.It is important to note that the random forest regression in the training model is only the provincial da
作者
王海灵
杨亚宁
康紫依
李淑娟
WANG Hailing;YANG Yaning;KANG Ziyi;LI Shujuan(School of Economics and Management,Xinjiang University,Urumqi 830047,China;Business School,Hohai University,Nanjing 211100,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2023年第8期2849-2858,共10页
Journal of Safety and Environment
基金
国家社会科学基金项目(20BGL015)
国家级大学生创新训练计划项目(S202110755064)。
关键词
公共安全
组合最优化
防疫
物流系统
机器学习
public safety
combinatorial programming
epidemic prevention
logistics system
machine learning
