Comparative Analysis of Machine Learning Techniques for Pneumonia Detection in Chest X-Ray Images

Lower respiratory infections, such as pneumonia, represent a major global health problem, ranking as the world's most deadly communicable disease after COVID-19. Identifying the causative organism type (bacterial or viral) is critical, as treatments differ significantly, and bacterial pneumonia tends to develop more severe outcomes than viral cases. Imaging diagnosis is essential but relies on specialists' subjective analysis. This study aimed to develop an automated system for pneumonia identification and classification using chest X-rays to improve diagnostic accuracy and reduce analysis time. We employed the public dataset Chest X-Ray Images (Pneumonia) from Kaggle, containing 5,856 X-ray images, to train three models: Convolutional Neural Network, Random Forest, and Support Vector Machine. Preprocessing included resizing (224 × 224 pixels), normalization, and data augmentation for class balancing. The Convolutional Neural Network-based DenseNet121 model achieved the highest performance, with 81% accuracy and an area under the curve of 84.2% in multiclass classification. These results demonstrate that accessible artificial intelligence techniques can effectively support imaging diagnosis, offering clinical decision-making tools with improved speed and precision.