Incorporating weighted block sampling techniques in epidemic curve classification.

Area of Science:

• Epidemiology
• Computational Biology
• Machine Learning

Background:

• Infectious disease outbreaks exhibit dynamic transmission rates, often peaking mid-epidemic.
• Individual-level models (ILMs) are crucial for understanding disease spread but are computationally intensive when using Bayesian MCMC frameworks.
• Standard random forests lack the temporal sensitivity needed for outbreak analysis, as they treat all data points equally.

Purpose of the Study:

• To develop a computationally efficient and accurate method for analyzing infectious disease outbreaks.
• To improve upon standard random forest models by incorporating temporal dynamics of disease transmission.
• To enhance real-time epidemic prediction and response planning.

Main Methods:

• Proposed a block-sampling-based random forest algorithm.
• Assigned higher weights to data blocks containing peak transmission times.
• Validated the model using simulated data and the 2001 UK foot-and-mouth disease outbreak data.

Main Results:

• The proposed weighted random forest model demonstrated superior performance compared to standard random forests.
• The method effectively identified and prioritized critical periods of the epidemic, specifically the peak transmission time.
• Achieved faster analysis times compared to traditional Bayesian MCMC methods.

Conclusions:

• The block-sampling-based random forest offers a computationally efficient and accurate approach for infectious disease outbreak analysis.
• This method enhances the prediction of epidemic generating models by focusing on critical temporal dynamics.
• The approach is suitable for real-time epidemic prediction and informing public health response strategies.