Identification of COVID-19 from Other Upper Respiratory Tract Infections Using Random Undersampling and LIME-based XAI Model

The urgent need for an accurate and fast diagnosis of upper respiratory tract infections (URTI) has been made clear by the COVID-19 pandemic. Machine learning algorithms have shown the potential to aid early diagnosis and appropriate patient management by distinguishing COVID-19 from other URTIs. By analyzing clinical and laboratory factors, such as symptom profiles, vital signs, and laboratory markers, machine learning algorithms can identify distinctive patterns that set COVID-19 apart from other URTIs. Healthcare workers might use such approaches to make well-informed decisions about patient triage, isolation, and treatment, which could result in improved clinical outcomes and slower disease transmission. To achieve this goal, the current work utilizes a publicly accessible dataset of symptoms and Explainable AI to extract the most significant aspects of the symptoms. The class imbalanced problem in the dataset has been eradicated by Random Under Sampling technique. In addition, classification algorithms have been used to determine the type of disease from patient’s symptoms. These algorithms included Decision Tree (DT), Random Forest (RF), Gradient Boosting (GB), Logistic Regression (LR), Naïve Bayes (NB), Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Extra Tree (ET), Bagging Classifier (BG) and Stacking Classifier. The results of the proposed work indicate that the stacking ensemble model got the highest 97.31% of classification accuracy and outperformed the previous work.