EpiVECS: exploring spatiotemporal epidemiological data using cluster embedding and interactive visualization.

Area of Science:

• Epidemiology
• Data Science
• Computational Biology

Background:

• Analyzing spatiotemporal data is crucial for descriptive epidemiology but remains challenging due to data complexity.
• Effective methods are needed to simplify the exploration of epidemiological time-series data for surveillance and hypothesis generation.

Purpose of the Study:

• To evaluate cluster embedding methods for spatially visualizing patterns in epidemiological time-series data.
• To compare the performance of different cluster embedding techniques using internal cluster validation metrics.
• To introduce EpiVECS, a novel tool for interactive exploration of cluster embedding results.

Main Methods:

• Utilized combined clustering and dimensionality reduction techniques, termed 'cluster embedding' methods.
• Compared k-means clustering-based approaches against self-organizing maps for epidemiological data.
• Employed internal cluster validation metrics to assess method performance.

Main Results:

• K-means clustering-based methods generally demonstrated superior performance compared to self-organizing maps on real-world epidemiological datasets.
• Identified specific cluster embedding techniques that excel in visualizing spatiotemporal disease patterns.
• Developed and validated the EpiVECS tool for user-friendly cluster embedding and visualization.

Conclusions:

• Cluster embedding methods, particularly those based on k-means, offer a promising approach to simplify spatiotemporal epidemiological data analysis.
• The EpiVECS tool provides an accessible, privacy-preserving platform for exploring complex disease patterns.
• This work facilitates enhanced epidemiological surveillance and hypothesis generation through improved data visualization.