Risk Assessment gives those who use the corus app an advantage to predict their risk for catching COVID-19. Using a machine learning algorithm, certain parameters such as the user's location, contact with COVID-19 positive individuals, pre-existing conditions, and breathing/coughing audio analysis, are taken into consideration and the risk associated is provided.
The proposed model is a weighted ensemble CNN (Convolutional Neural Network). CNNs are used for feature extraction, and our goal is to reduce the initial array to a 1D-vector which retains the features of the audio. In this procedure, three total CNNs will be used, and their predictions will be averaged (hence the name weighted ensemble approach). The AUC curve, overall accuracy, and confusion matrix will be used to interpret the results.