Elephant Herding Optimization algorithm based Maximal Relevancy Feature selection in Chronic Obstructive Pulmonary Disease Prediction

Abstract

Chronic Obstructive Pulmonary Disease (COPD) is a progressive respiratory disorder and a major global health challenge, making early and
accurate prediction critical for improving patient outcomes. However, existing COPD prediction models face limitations in feature selection, as
conventional filter, wrapper, and hybrid techniques often lead to redundant features, high computational cost, and poor handling of nonlinear
interactions. Likewise, traditional optimization algorithms are prone to premature convergence, which restricts their ability to identify the most
relevant predictors. This creates a research gap in developing efficient feature selection methods that achieve maximal relevancy, minimal
redundancy, and strong convergence for robust prediction. To address this, the proposed study introduces an Elephant Herding Optimization
(EHO - FSSO) with Maximal Relevancy Feature selection framework for COPD prediction. Inspired by elephant social behaviour, EHO enhances
global search and avoids local optima, while the maximal relevancy criterion ensures the selection of compact yet informative feature subsets.
The selected features are validated using three machine learning classifiers such as Random Forest, Decision Tree, and Artificial Neural
Network—providing a comprehensive assessment of predictive performance. This approach improves accuracy, reduces computational
complexity, and ensures interpretability of results. The expected contributions include an efficient EHO-driven feature selection strategy for
COPD, validated performance across multiple classifiers, and a scalable, clinically meaningful framework for early disease prediction and
decision support.