Pneumonia Detection from CXR using Adaptive Elephant Herd Optimization and Python Rectilinear Locomotion Strategy

Abstract Pneumonia poses a major global health challenge, impacting around 450 million individuals annually. This highlights the urgent need for computerized pneumonia detection using chest X-ray (CXR) images. Today, deep learning (DL) models are used to analyze CXR images for accurate pneumonia diagnosis. Hyperparameters significantly influence the effectiveness of DL models in disease prediction. Effective tuning of these parameters is essential to develop robust, accurate, and efficient predictive models. This paper explores several baseline hyperparameter optimization techniques for tuning them in the pneumonia detection from CXR images using convolutional neural networks (CNNs). Additionally, an adaptive elephant herd optimization (AEHO) using the Python rectilinear locomotion strategy (PRLS) is proposed in this paper to enhance disease prediction models. The proposed AEHO-PRLS model achieved 96.57 % accuracy and outperformed the baseline models in accuracy and reliability of disease prediction models.