Evaluation of Mutual Information and Feature Selection for SARS-CoV-2 Respiratory Infection

Area of Science:

• Computational epidemiology
• Machine learning in public health
• Infectious disease modeling

Background:

• Accurate forecasting of SARS-CoV-2 spread is crucial for effective public health planning and resource allocation.
• Existing models may not fully capture the stochastic nature of virus transmission or handle data uncertainties.
• Understanding country-wise pandemic dynamics requires advanced analytical tools.

Purpose of the Study:

• To develop a predictive machine learning model for SARS-CoV-2 respiratory infections.
• To enhance prediction accuracy through feature selection and data visualization.
• To analyze country-wise pandemic dynamics using machine learning.

Main Methods:

• Utilized stochastic regression (SR) to model virus transmission dynamics and data uncertainties.
• Employed feature selection techniques to identify key predictive variables.
• Applied neighbor embedding (NE) and Sammon mapping (SM) for visualizing high-dimensional data.
• Evaluated algorithms including neural networks (NN), decision trees (DT), and random forests (RF) with Adam optimizer (AD) and hyperparameters (HP).

Main Results:

• A novel orchestration combining pre-processed data with ADHPSRNESM demonstrated high prediction accuracy.
• Feature selection significantly improved the precision of SARS-CoV-2 spread predictions.
• Data visualization techniques provided better interpretation of underlying pandemic patterns.

Conclusions:

• The developed machine learning model offers a precise forecasting tool for SARS-CoV-2 respiratory infections.
• Informed decision-making for policymakers and healthcare professionals is enhanced by accurate predictions.
• The study contributes to improved management and control strategies for the SARS-CoV-2 pandemic.