DESIGN OF CLASSIFICATION MODEL FOR COVID-19 DETECTION THROUGH RECORDING OF COUGH USING XGBOOST CLASSIFIER ALGORITHM

• Main Author: Elida Irawati, Mesayu
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/55927
• Institution: Institut Teknologi Bandung

Early detection of viral infections is crucial in dealing with the rate of spread of the outbreak. People exposed to Covid-19 can immediately carry out self-isolation and home care so that transmission does not spread and endanger the health of those around them, who are more susceptible to exposure due to a weak immune system, especially someone with a history of comorbidities. Currently, the gold standard for the diagnosis of COVID-19 is RT-PCR (Reverse Transcription Polymerase Chain Reaction). However, this test has limitations such as the scarcity of the PCR device, examiners' skills in performing procedures, and risks to breaking social distancing protocol. Other Covid-19 screening tools such as the rapid antigen, rapid antibody test, and GeNose share the same limitations. Therefore, model development is needed to realize the Covid-19 screening test that is free, noninvasive, and can be available anywhere to overcome those limitations. A smartphone application has the potential for the Covid-19 pre-screening tool available for a wider community. Covid-19 is known to attack the respiratory system. Testing the cough sound can be done as early detection to determine if someone exposes to Covid-19. In this study, We developed a machine learning model to classify Covid-19 through recorded cough sounds. MFCC features extracted from cough sound. Then, the XGBoost Classifier algorithm labeled the cough sound based on these features. We trained the algorithm against Virufy and Corwara databases. Results showed that the accuracy of the classification model reached 86%.