Enhancing COVID-19 Classification Accuracy with a Hybrid SVM-LR Model.

Area of Science:

• Machine Learning
• Biostatistics
• Epidemiology

Background:

• Support Vector Machine (SVM) and Logistic Regression (LR) are established classification algorithms.
• Recent advancements like bagging and ensemble methods have enhanced SVM and LR capabilities.
• Existing comparisons between SVM and LR predate these modern improvements.

Purpose of the Study:

• To propose and evaluate a novel hybrid model integrating SVM and LR.
• To assess the hybrid model's performance in predicting small events per variable (EPV).
• To compare the hybrid model against standalone SVM and LR using real-world pandemic data.

Main Methods:

• Development of a hybrid classification model combining SVM and LR.
• Evaluation of the hybrid, SVM, and LR models across various EPV values.
• Utilized COVID-19 epidemiological data from December 2019 to May 2020 (WHO).

Main Results:

• The hybrid SVM-LR model demonstrated superior classification performance.
• Outperformed standalone SVM and LR in accuracy, Mean Squared Error (MSE), and Root Mean Squared Error (RMSE).
• Consistent performance improvements observed across different EPV levels.

Conclusions:

• The proposed hybrid model offers enhanced predictive capabilities for epidemiological data.
• This approach is valuable for public health authorities in managing future pandemics.
• The hybrid model provides a robust tool for analyzing events with limited variables.