Nature inspired optimization model for classification and severity prediction in COVID-19 clinical dataset

Area of Science:

• Medical Informatics
• Computational Biology
• Machine Learning in Healthcare

Background:

• The emergence of mutated COVID-19 strains necessitates rapid diagnosis and severity assessment.
• Current diagnostic methods like RT-PCR provide binary results, but predicting disease severity is crucial for patient management.
• Machine learning offers a powerful approach to analyze complex clinical data for improved diagnostic and prognostic tools.

Purpose of the Study:

• To develop and evaluate machine learning models for the accurate screening and severity prediction of COVID-19.
• To identify an optimal subset of clinical features for model development using an artificial bee colony optimization algorithm.
• To aid medical practitioners in timely and effective patient care decisions.

Main Methods:

• Analysis of a large clinical dataset (65,000 patient records, 26 features).
• Feature selection using a variant of the artificial bee colony optimization algorithm.
• Development of a binary classifier using Support Vector Machine (SVM) for COVID-19 screening.
• Implementation of a logistic regression model for predicting disease severity (mild, moderate, severe).

Main Results:

• The Support Vector Machine model achieved a high accuracy of 96% for COVID-19 patient screening.
• The logistic regression model successfully predicted three levels of disease severity.
• The binary classifier demonstrated strong efficiency with precision-recall values of 0.96 and 0.97.
• The severity prediction model showed high accuracy, with 96.0% for mild/moderate cases and 85.0% for severe cases.

Conclusions:

• Machine learning models, particularly SVM and logistic regression, are effective tools for COVID-19 diagnosis and severity prediction.
• Optimized feature selection enhances model performance.
• These predictive models can significantly assist clinicians in determining appropriate patient treatment and resource allocation.