Convolutional Neural Network Based Identification of Respiratory Disease (CNN-IRD)

Abstract

Coronavirus is a type of virus that can cause Respiratory Disease (RD) in people. The World Health Organization (WHO) states that signs and symptoms in mild cases include dry throat, fever, nasal secretions, shortness of breath, fever, and malaise. The disease is more dangerous than viruses and can cause serious illness. Although many researchers have tried various techniques for classifying RD patients, it is essential to identify the critical features before applying machine learning methods for classification to save time and cost. To this end, this paper proposes a Convolutional Neural Network (CNN) based Identification of RD patients (CNN-IRD) caused by Coronavirus and divides them into two classes, i.e., C19+ve and C19-ve. First, we apply the binarization technique to preprocess data into useful information. Second, we identify the significant features using Linear Discriminant Analysis (LDA). Finally, we train a deep learning classifier (CNN) with two publicly available datasets. The evaluation results suggest that CNN yields other classifiers in predicting RD patients. The performance improvements of CNN-IRD in accuracy, precision, recall, and f-measure with both datasets are (7.92%, 5.35%, 16.92%, and 11.22%) and (4.04%, 9.08%, 25.18%, and 17.25%), respectively.