Prediction of Covid-19 Using Ensemble based Machine Learning Approach

Abstract

Abstract- The viral infection of COVID-19 has been wreaking havoc on earth since December 2019. It has affected every aspect of human life. This dangerous disease took millions of lives and made even larger number of people sick with dreadful symptoms. To stop the spread of this fatal infection reliable COVID-19 screening is crucial at an early stage. Some studies have been conducted which depict the importance of using routine blood test for initial screening of COVID-19 positive patients. In this study we employed several machine learning techniques to predict COVID-19 using complete blood count. Variants of the different classifiers are generated by tuning their parameters to get better accuracy. The algorithms that are tuned and optimized include " Support Vector Machine, K Nearest Neighbors, Multi-Layer Perceptron, Random Forest, Decision Tree and Naive Bayes". These variants are used to detect COVID-19. Variants which depict higher accuracy are chosen from each classifier family. Ensemble learning is used to integrate the highest performing variants. Decision Tree variant is chosen for ensemble as it performed better than the other classifier variants. “Accuracy, Recall, Precision and F-Measure" are the performance measures used.

Keywords: - Machine learning, COVID-19 Prediction, Ensemble learning, Classifier Variants, CBC Test.