Epidemiologically and Socio-economically Optimal Policies via Bayesian Optimization

Area of Science:

• Epidemiology
• Computational Science
• Public Health Policy

Background:

• Mass public quarantining (lock-downs) are non-pharmaceutical interventions to control disease spread.
• Balancing public health and socio-economic impacts of lock-downs is a critical challenge.
• Existing methods may not optimally balance these competing factors.

Purpose of the Study:

• To present ESOP (Epidemiologically and Socio-economically Optimal Policies), a novel machine learning approach.
• To develop an AI-driven method for optimizing lock-down schedules.
• To balance public health benefits against socio-economic costs.

Main Methods:

• Utilized active machine learning techniques, specifically Bayesian optimization.
• Developed ESOP to interact with epidemiological models in a black-box manner.
• Introduced VIPER (Virus-Individual-Policy-EnviRonment), a stochastic agent-based simulator for case studies.

Main Results:

• Demonstrated the utility of ESOP through case studies using the VIPER simulator.
• ESOP can generate multi-phase lock-down schedules.
• The approach effectively balances epidemiological and socio-economic considerations.

Conclusions:

• ESOP offers a flexible and powerful tool for optimizing public health interventions.
• This AI-driven strategy can inform more effective and balanced lock-down policies.
• The method is adaptable to various epidemiological models and scenarios.