流空间邻近关系约束下的流行病分布空间异常探测方法

In view of the limitations of existing epidemic spatial anomaly detection methods in comprehensively detecting potential spatial anomalies caused by multiple factors, this paper proposes an epidemic distribution spatial anomaly detection method under the constraint relationship of flow space proximity. Firstly, based on the geographical detector, the epidemic thematic attributes that have a significant correlation with the outflow intensity factor of the population in the transmission center are identified; then, the flow space weight matrix is adaptively constructed based on the flow space proximity relationship measurement; finally, the spatial local change gradient variable of epidemic attributes is constructed to characterize the epidemic situation characteristics of spatial units, and an improved global and local Moran's I is proposed to achieve the statistical discrimination of the epidemic distribution pattern in flow space and the detection of local spatial anomaly areas. The example of the COVID - 19 epidemic verifies that compared with the existing Euclidean space anomaly detection methods, the method in this paper can effectively identify the epidemic distribution spatial anomaly areas caused by multiple types of potential factors other than the cross - regional flow of the population during the epidemic development process, which is helpful to support the phased, zoned and graded precise prevention and control of the epidemic.

针对现有流行病空间异常探测方法在全面探测多因素导致的潜在空间异常方面的局限性,本文提出一种流空间邻近约束关系下的流行病分布空间异常探测方法。首先,基于地理探测器识别与传播中心人群流出强度因素具有显著关联关系的疫情专题属性;然后,基于流空间邻近关系度量自适应构建流空间权重矩阵;最后,构造疫情属性空间局部变化梯度变量刻画空间单元疫情态势特征,提出改进的全局和局部莫兰指数(Moran's I)实现流空间疫情分布模式的统计判别与局部空间异常区域探测。新型冠状病毒肺炎(COVID-19)疫情的实例,验证了本文方法相比现有欧氏空间异常探测方法,能够有效识别疫情发展过程中除人群跨区域流动之外的多类潜在因素导致的疫情分布空间异常区域,有助于支持对疫情分阶段的分区分级精准防控。