This study aims to develop a computer-based clinical decision support system that will help clinicians and healthcare personnel to make an early and correct decision to prevent the patient from nontransmissible respiratory diseases. The main contribution of this study is to analyze, investigate, and extraction of the useful feature of pathological respiration and Classification of Crackle and Wheeze from recorded lungs sound by using machine learning techniques. In the particular spectrogram, Time-frequency and Mel-Frequency cepstral coefficient (MFCC) technique is applied for feature analysis and data conversion into a format that can be useful for feature extraction and training models. PCA dimensional reduction technique is used to reduce the dimensionality of the extracted feature. In order to apply various machine learning techniques a widely used dataset freely available dataset ICBHI-2017 is used. The respiratory lungs sound is comprised of 126 patients with 920 Chest sound annotations that include adventitious sounds such as “Crackle” and “Wheeze”. Machine learning algorithms such as MusicANN, VGGish, and OpenL3 were applied for testing the better accuracy of the classification model. The accuracy of the utilized classifier with the extracted feature set is determined as 72%, 81%, and 69% respectively.