[论文解读] Association Learning Between the COVID-19 Infections and Global Demographic Characteristics Using the Class Rule Mining and Pattern Matching

Wasiq Khan, Abrar Hussain+3 more|arXiv (Cornell University)|Sep 27, 2020

COVID-19 epidemiological studies参考文献 32被引用 4

一句话总结

本研究提出一种结合类别规则挖掘与模式匹配的智能模型，以揭示全球人口统计特征与新冠肺炎严重程度之间的多维关联。基于截至2020年8月20日的数据，运用自组织映射、类别关联规则及统计分析，识别出显著关联，特别是与女性吸烟者相关的关联，表明人口统计分布对各地区疾病进展和严重程度具有显著影响。

ABSTRACT

Over 26 million cases have been confirmed worldwide (by 20 August 2020) since the Coronavirus disease (COIVD_19) outbreak in December 2019. Research studies have been addressing diverse aspects in relation to COVID_19 including potential symptoms, predictive tools and specifically, correlations with various demographic attributes. However, very limited work is performed towards the modelling of complex associations between the combined demographic attributes and varying nature of the COVID_19 infections across the globe. Investigating the underlying disease associations with the combined demographical characteristics might help in comprehensive analysis this devastating disease as well as contribute to its effective management. In this study, we present an intelligent model to investigate the multi-dimensional associations between the potentially relevant demographic attributes and the COVID_19 severity levels across the globe. We gather multiple demographic attributes and COVID_19 infection data (by 20 August 2020) from various reliable sources, which is then fed-into pattern matching algorithms that include self-organizing maps, class association rules and statistical approaches, to identify the significant associations within the processed dataset. Statistical results and the experts report indicate strong associations between the COVID_19 severity levels and measures of certain demographic attributes such as female smokers, when combined together with other attributes. These results strongly suggest that the mechanism underlying COVID_19 infection severity is associated to distribution of the certain demographic attributes within different regions of the world. The outcomes will aid the understanding of the dynamics of disease spread and its progression that might in turn help the policy makers and the society, in better understanding and management of the disease.

研究动机与目标

• 建立全球各地区合并人口统计特征与新冠肺炎感染严重程度之间复杂关联的模型。
• 解决关于多维人口统计因素对疾病进展和严重程度影响的研究不足问题。
• 通过识别与严重结局相关的高风险人口统计聚类，支持公共卫生政策制定。
• 将模式匹配技术与统计验证相结合，实现稳健的关联发现。

提出的方法

• 从可靠来源收集截至2020年8月20日的全球人口统计与新冠肺炎感染数据。
• 应用自组织映射检测高维人口统计与感染数据中的潜在模式。
• 采用类别关联规则挖掘技术，提取人口统计特征与新冠肺炎严重程度等级之间的显著关系。
• 整合统计分析以验证所发现关联的显著性。
• 结合专家报告，评估研究发现的临床与流行病学合理性。

实验结果

研究问题

• RQ1 哪些人口统计特征组合在全球范围内与更高的新冠肺炎严重程度水平表现出强关联？
• RQ2 各地区的人口统计分布如何与疾病进展和严重程度的差异相关联？
• RQ3 哪些特定的人口统计分组（如女性吸烟者）表现出对严重结局的显著预测能力？
• RQ4 模式匹配与规则挖掘技术在多维健康数据中揭示非显而易见关联的程度如何？

主要发现

• 发现女性吸烟者与新冠肺炎严重程度之间存在强关联，提示其为显著的人口统计风险因素。
• 自组织映射与类别关联规则的整合揭示了人口统计聚类与疾病结局之间的复杂非线性关系。
• 统计验证与专家评审证实了所识别关联的显著性与合理性。
• 该模型表明，各地区的人口统计分布与疾病严重程度模式的差异密切相关。
• 研究结果表明，性别、吸烟行为与区域分布等人口统计特征共同影响疾病传播与严重程度的动力学。