Optimal Feature Selection with Modified Dynamic Crow Search Algorithm for IoT-Spiro Based Chronic Obstructive Pulmonary Disease Prediction

Chronic Obstructive Pulmonary Disease (COPD) is a major global health issue, which requires early diagnosis in order to ensure effective treatment and management. With the evolution of IoT-based healthcare systems, large-scale sensor data is now used for clinical analysis, which makes automated diagnostic methods increasingly important. However, datasets contain irrelevant or redundant features, which impacts the performance of classifiers. To address this research issue, a Modified Dynamic Crow Search Algorithm (mDCSA) has been proposed for optimal feature selection in COPD prediction using IoT-Spiro data. The mDCSA incorporates adaptive parameter control and binary encoding to efficiently balance exploration and exploitation during the search process. The selected optimal feature set was evaluated using various machine learning classifiers. Experimental results demonstrated that mDCSA significantly improves classification accuracy, with Support Vector Machine (SVM) achieving the highest overall performance across several key evaluation metrics.