A Hybrid Approach for COVID-19 Pneumonia Detection and Classification using Machine and Deep Learning Techniques

Abstract

Pneumonia, including COVID-19 pneumonia, remains a significant global health challenge, with early and accurate diagnosis being essential for effective management. Traditional diagnostic methods often struggle with low accuracy and limited adaptability to evolving disease patterns. The COVID-19 pandemic highlighted the need for more robust, adaptive diagnostic approaches.This study develops an integrated machine learning (ML) framework for classifying COVID-19 pneumonia using Random Forest, AdaBoost, XGBoost, and Convolutional Neural Networks (CNNs). The research aims to optimize these algorithms, improve data preprocessing, and evaluate their performance compared to traditional methods.The proposed framework consists of four phases: (1) Dataset Acquisition: Utilization of a GitHub dataset, processed in Python and Anaconda Jupyter Notebook; (2) Data Processing and Analysis: Histograms and scatter plots for dataset preprocessing; (3) Model Application and Optimization: Random Forest and AdaBoost for ensemble learning, XGBoost with data augmentation for enhanced accuracy, and CNNs for extracting intricate patterns from X-ray images; and (4) Performance Evaluation: A comparative analysis of the integrated model against traditional methods to assess improvements in accuracy, reliability, and adaptability. Results demonstrate that XGBoost achieves the highest performance, with an accuracy of 87.35%, precision of 89.58%, recall of 87.22%, and an F1-score of 88.39%. CNN performs well in precision (88.60%), while AdaBoost provides balanced precision and recall. The integrated ML framework significantly enhances pneumonia classification accuracy, offering a promising tool for COVID-19 diagnostics. These findings emphasize the importance of model selection to optimize both precision and recall, potentially improving clinical decision-making and patient outcomes.

Keywords: COVID-19 pneumonia, XGBoost, classification, chest X-rays, prediction