Early diagnosis is the need of the hour in the treatment of respiratory-related health conditions. This study presents a novel method for monitoring respiratory disorders by applying a Least Absolute Shrinkage and Selection Operator (LASSO) regression model to Photoplethysmography (PPG) signals. By analyzing respiratory variations in the PPG waveform, the partial pressure of carbon dioxide (PCO₂) signal is extracted to monitor breathing patterns. The PCO₂ signal provides critical insights into respiratory dynamics, enabling the identification of irregular breathing rates and airflow obstructions. Using LASSO regression, the most relevant features from the PCO₂ signals are selected, reducing dimensionality and improving prediction accuracy. The proposed approach offers a cost-effective and non-invasive solution for evaluating respiratory health, making it suitable for both clinical and non-clinical settings. A comprehensive performance analysis demonstrates the efficacy of the LASSO regression-based method in diagnosing respiratory conditions. To evaluate its performance, five machine learning classifiers were employed: Linear Regression, Bayesian Linear Discriminant Analysis (BLDA), k-Nearest Neighbors (k-NN) with weighted voting, Expectation-Maximization (EM) with Logistic Regression, and Elephant Search Optimization (ESO). The results highlight the potential of this approach to improve healthcare by enabling early detection and management of respiratory disorders. The Elephant Search Optimization, combined with LASSO regression for dimensionality reduction, achieves 95.12% accuracy value, 95% F1 score, 0.90% MCC value, 4.87% error rate, 90.47% in Jaccard metrics, and 90% CSI.

Elephant Search Optimization (ESO) LASSO PCO2 Bayesian LDA PPG