Spatial Data Science of COVID-19 Data

Siyuan Shang, Carson K. Leung, Yubo Chen, Adam G.m. Pazdor2020 · DOI

Abstract

Huge amounts of big data can be generated and collected from a wide variety of rich data sources. Embedded in these big data are useful information and valuable knowledge. An example is healthcare and epidemiological data such as data related to patients who suffered from viral diseases like the coronavirus disease 2019 (COVID-19). Knowledge discovered from these epidemiological data via data science helps researchers, epidemiologists and policy makers to get a better understanding of the disease, which may inspire them to come up ways to detect, control and combat the disease. In this paper, we present a spatial data science system for analyzing big COVID-19 epidemiological data, with focus on the spatial data analytics among different geographic locations. The system helps users to get a better understanding of information about the confirmed cases of COVID-19. Evaluation results show the benefits of our system in spatial data analytics of big COVID-19 data.

Introduction

The COVID-19 pandemic generated an unprecedented volume of epidemiological data. While public dashboards provided immediate visibility into case counts and mortality rates, they often lacked the depth required to understand complex spatial patterns. Big data in this context is defined by the four V’s—Volume, Velocity, Variety, and Veracity—presenting significant challenges for extracting actionable knowledge. This paper presents a spatial data science system designed to analyze these complexities, moving beyond surface-level metrics to uncover how symptoms, transmission methods, and outcomes vary across geographic scales.

Methodology: A Multi-Stage Spatial Pipeline

The proposed system utilizes a multi-stage pipeline to transform raw, often incomplete, epidemiological reports into structured knowledge.

Data Integration and Preprocessing

A key challenge in pandemic data is the prevalence of missing values (NULLs). Rather than discarding these records, which could introduce selection bias, the system retains them. Data is further anonymized and standardized through "binning"—grouping ages into decades and dates into weekly windows—to balance privacy with statistical utility.

Spatial Hierarchy and Mining

Central to the system is a hierarchical data structure that allows for multi-level analysis. Users can "zoom" from local facilities up to global continents.

The spatial hierarchy tree used to organize epidemiological data across different geographic scales.

The system applies frequent pattern mining to identify co-occurring attributes (e.g., specific symptom clusters) and contrast mining to compare these patterns across different regions. This reveals unique regional characteristics that might be obscured in national-level aggregates.

Datasets and Evaluation

The system was evaluated using integrated data from the World Health Organization (WHO), Statistics Canada, and the Public Health Agency of Canada (PHAC). The primary study focused on a dataset of 209,811 Canadian cases, incorporating detailed attributes such as symptoms, transmission methods, and clinical outcomes.

Results and Discussion

The spatial analysis revealed several critical insights that traditional reporting might overlook.

Relative vs. Absolute Risk

While absolute case counts are often the focus of media reporting, relative risk provides a more accurate picture of regional impact. For instance, although Ontario and Quebec both had high absolute numbers, spatial analysis showed that Quebec's infection rate relative to its population was significantly higher than the national average.

Comparison of cumulative COVID-19 cases and infection rates across Canadian national regions.

Symptom and Recovery Patterns

Mining revealed that symptoms often cluster; for example, the pattern {cough, recovered} appeared in over 74% of a specific symptomatic subset. Furthermore, recovery rates showed significant spatial variance. The Atlantic provinces maintained a recovery rate of approximately 93.4%, compared to roughly 71.7% in the Prairies + NWT region, suggesting that geographic location and regional policy significantly influenced clinical outcomes.

Conclusion

This spatial data science framework demonstrates the power of hierarchical mining in understanding large-scale health crises. By partitioning data spatially, the system achieves the scalability necessary for big data while providing the granularity needed for local policy decisions. Future work will focus on integrating visual analytics to make these complex patterns more accessible to non-technical stakeholders and applying the framework to other domains like urban planning.

Summary updated

Abstract

Huge amounts of big data can be generated and collected from a wide variety of rich data sources. Embedded in these big data are useful information and valuable knowledge. An example is healthcare and epidemiological data such as data related to patients who suffered from viral diseases like the coronavirus disease 2019 (COVID-19). Knowledge discovered from these epidemiological data via data science helps researchers, epidemiologists and policy makers to get a better understanding of the disease, which may inspire them to come up ways to detect, control and combat the disease. In this paper, we present a spatial data science system for analyzing big COVID-19 epidemiological data, with focus on the spatial data analytics among different geographic locations. The system helps users to get a better understanding of information about the confirmed cases of COVID-19. Evaluation results show the benefits of our system in spatial data analytics of big COVID-19 data.
BibTeX
@misc{shang2020spatial,
  author = {Siyuan Shang and Carson K. Leung and Yubo Chen and Adam G.M. Pazdor},
  title = {{Spatial Data Science of COVID-19 Data}},
  year = {2020},
  doi = {10.1109/hpcc-smartcity-dss50907.2020.00177},
}
Figure