APPLICATION OF CONVOLUTIONAL NEURAL NETWORK AND DECISION TREE IN THE DIAGNOSIS OF THE CHRONIC OBSTRUCTIVE PULMONARY DISEASE WITH SOUND DATA

• Main Author: Alexander Liong, Steven
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/72951
• Institution: Institut Teknologi Bandung

Technological advances, particularly in machine learning and deep learning, have played an important role in addressing challenges in medicine. With limited human resources in the medical field and traditional methods, the implementation of this technology can be a solution. Chronic Obstructive Pulmonary Disease or COPD for short is one of the 10 deadliest diseases in the world. Early diagnosis and timely treatment are very important in dealing with COPD. This study aims to determine the best method between the Convolutional Neural Network (CNN) model method and the Decision Tree in diagnosing COPD. The data used is the 2017 ICBHI breathing sound data, which consists of 920 patient breathing files along with a description of the file. Model evaluation was performed with accuracy, precision, recovery, F1 score, and confusion matrix. In the experiments conducted, the authors utilized various features from the Pandas, Numpy, Librosa, Tensorflow, and Scikit-Learn machine learning libraries. In addition, sound extraction methods such as MFCC, Chroma-STFT, and Mel-Spectrogram are used.