A Diagnostic Study on Prediction of Covid-19 by Symptoms Using Machine Learning

• Published in: 2022 International Conference on Electronics and Renewable Systems (ICEARS) pp. 1416 - 1421
• Main Authors: Devi, E. Anna, Athappan, V., Rajendran, Rahul R, Devi, E. Ahila, Emayavaramban, G., Sriragavi, S., Sivaramkrishnan, M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.03.2022
• Subjects: COVID-19; Decision trees; Hybrid Model; Machine learning algorithms; Measurement; Performance Metrics; Pre-process; Predictive models; Renewable energy sources; ROC Curve; Training
• DOI: 10.1109/ICEARS53579.2022.9752301

Effective screening helps for quick and accurate detection of COVID-19 and it also decreases the burden on the healthcare system. Prediction models with numerous criteria have been developed to estimate the probability of infection. These are designed to assist medical workers across the world in triaging victi ms, especially in places with limited medical resources. For predicting the COVID-19 using symptoms, the dataset is taken from the website of the Israeli Ministry of Health. The dataset contains 9 attributes and 2,78,848 samples. The raw dataset is cleaned using pre-processing techniques. The Machine learning algorithms like Random Forest, K Nearest Neighbor, Decision Tree, and hybrid Random Forest, K Nearest Neighbor, and Decision Tree are applied on the 1,95,194 samples to identify the model. The predicted model is tested on 83,654 samples to ensure the quality of the designed model. The performance metrics like ROC [Receiver Operating Characteristic] curve, True Positive and Negative Rate, False Positive and Negative Rate, Positive and Negative Predictive Value, and Accuracy are applied to check the model. From the evaluation result, the proposed hybrid model gives high accuracy of 98.97% . The proposed technique might be utilized to priorities COVID-19 screening when testing capabilities are constrained., among several other things.