一种优化的Logistic模型的流行病分类方法

Epidemiological classification method based on optimized logistic model

对流行病的分类进行定量约束,为了确定某感染病是否为大流行病,分别搜集了以H1N1为代表的“大流行病”重灾区样本和以SARS为代表的“流行病”重灾区样本,并定义每次流行病的“重灾区”。其次分别选取感染率、病死率、医疗条件、人口密度等指标作为解释变量建立二分类Logistic模型,通过得到的被解释变量的预测值(即该病是否为大流行病的概率)来判断该病是否为大流行病。最后,由于疾病流行性问题的复杂性,在模型中依次引入平方项重新回归,通过定义不同的训练集和测试集,得到最优的回归方程,最终得到将病死率的平方项引入回归模型中,模型分类的准确率更高。

In this paper,we quantitatively restrict the classification of epidemics.In order to determine whether an infectious disease is a pandemic,we collected samples of"pandemic"severely affected areas represented by H1N1 and"epidemic"severely affected areas represented by SARS.Define the"heavy disaster area"for each epidemic sample.Secondly,select the infection rate,mortality,medical condition,population density and other indicators as explanatory variables to establish a binary classification logistic model,and judge the disease by the predicted value of the explained variable(that is,the probability of whether the disease is a pandemic)whether it is a pandemic.Finally,due to the complexity of the disease epidemic problem,we re-regressed by introducing square terms in the model in turn,by defining different training sets and test sets,we got the optimal regression equation,and finally got the square term of the mortality rate into the regression model in the model classification accuracy is higher.