Respiratory diseases such as chronic obstructive pulmonary disease (COPD) and pulmonary diseases in general represent some of the most pervasive health threats globally, so there is a need to develop effective diagnostic systems. Therefore, identifying COPD at an early stage may help define early therapeutic approaches and individual patient care for this condition. The epidemiological data of COPD and its grave consequences for the health of patients show the importance of the development of effective diagnostic tools. The various conventional diagnostic approaches may not possess the detail needed to identify early-stage COPD or the difference between it and other respiratory diseases. This research seeks to create an adaptive and accurate model for assessing pulmonary audio data to diagnose COPD early. The objective is to increase diagnostic accuracy and employ modern approaches to machine learning algorithms and feature extraction. In this study, the feature extraction applied in Pulmonary sound recordings is Mel-frequency cepstral coefficients (MFCCs). Due to issues with dimensionality and computational complexity, the relevant features are selected using Forward Feature Selection (FFS). The classification approach synthesizes two methods, support vector machines, and k-nearest neighbors, to reveal intricate patterns and boundaries in the data. The COPD disease data set is the basis for all modeling and testing. The adopted SVM-KNN fusion model integrated in Python proved to deliver a high level of performance with an accuracy of 94 %. Again, this degree of accuracy attests to the model's effectiveness in distinguishing between healthy and COPD-affected lungs. The developed framework considerably improves the identification of patients' COPD and respiratory illness risk assessment.