In an era of advanced technology development, the early diagnosis of multivariate respiratory diseases is highly popular due to the unavailability of efficient solutions in the biomedical field. To solve lung respiratory diseases, an innovative structure is put forth to enable precise prediction and early diagnosis of multivariate respiratory Chronic Obstructive Pulmonary Diseases (COPD) using deep learning techniques. The proposed structure performs lung sound data preprocessing to extract the features using noise filtering techniques. After pre-processing, the framework extracts the characteristics that discriminate from the lung sound to train the model based on extracted features. The proposed framework extracts various features such as constant-Q transform, short-time Fourier transform, and Mel frequency central coefficients, the model is constructed for accurate prediction of respiratory diseases using a convolutional neural network (CNN). The effectiveness of the suggested fra has been evaluated on the open lung sound database in different benchmark settings to measure the different statistical correlations with existing methods for a prompt diagnosis and precise forecasting of COPD diseases achieving an accuracy of 95.48% with a sensitivity of 89.0% and specificity of 98.0% in predicting conditions like COPD, healthy, URTI, Bronchiolitis and pneumonia.