Spatio-temporal mixture process estimation to detect dynamical changes in population.

Area of Science:

• Ecology
• Public Health
• Data Science

Background:

• Population monitoring is crucial for public health and ecology but presents significant challenges.
• Existing methods struggle with accurately modeling dynamic population distributions in space and time.

Purpose of the Study:

• To develop an advanced method for modeling and monitoring population distributions across space and time.
• To create an alert system capable of detecting spatio-temporal data changes.

Main Methods:

• A novel Expectation-Maximization (EM) algorithm was developed to simultaneously estimate cluster numbers and parameters.
• The EM algorithm was integrated with a temporal statistical model to form a spatio-temporal mixture process (STMP).
• The STMP was tested on simulated datasets and validated using real-world COVID-19 patient data.

Main Results:

• The proposed EM algorithm demonstrated improved performance in estimating cluster parameters compared to existing methods.
• The STMP successfully modeled evolving population distributions and detected significant epidemic changes in COVID-19 data.
• The method accurately captured various dynamic behaviors of population distributions in synthetic datasets.

Conclusions:

• The developed spatio-temporal mixture process (STMP) provides a robust framework for monitoring dynamic population changes.
• This approach enhances the capability for early detection of critical shifts in spatio-temporal data, with applications in public health and ecology.
• The STMP pipeline offers a validated tool for analyzing evolving real-world data and identifying epidemiological trends.