Random-Forest-Bagging Broad Learning System With Applications for COVID-19 Pandemic

Area of Science:

• Epidemiology and Public Health
• Data Science and Machine Learning
• Computational Biology

Background:

• The COVID-19 pandemic has caused a global health crisis, necessitating accurate forecasting.
• Analysis and prediction of the COVID-19 pandemic have gained significant worldwide attention.
• Existing forecasting models may not fully capture the complex dynamics of the pandemic.

Purpose of the Study:

• To develop and evaluate a novel machine learning model for predicting COVID-19 pandemic trends.
• To compare the performance of the proposed model against various established machine learning algorithms.
• To leverage a comprehensive dataset encompassing pandemic, testing, economic, demographic, and geographic factors.

Main Methods:

• A large dataset of COVID-19 information from 184 countries and 1241 areas (December 2019–September 2020) was compiled.
• A random forest (RF) algorithm was used for feature selection.
• A random-forest-bagging broad learning system (RF-Bagging-BLS) was developed and applied for forecasting.

Main Results:

• The RF-Bagging-BLS model demonstrated superior forecasting performance compared to benchmark models.
• Key performance metrics including RMSE, R-squared, adjusted R-squared, MAD, and MAPE indicated the model's accuracy.
• The proposed model exhibited better predictive power than linear regression, KNN, decision tree, Ada, RF, GBDT, SVR, ETs, CatBoost, LightGBM, XGBoost, and BLS.

Conclusions:

• The RF-Bagging-BLS model offers a robust and accurate approach for COVID-19 pandemic forecasting.
• This machine learning approach provides valuable insights for public health decision-making and resource allocation.
• The study highlights the potential of advanced machine learning techniques in managing global health crises.