Early stage prediction of COVID-19 Using machine learning model

Abstract

The healthcare sector has traditionally been an early use of technological progress and has achieved significant advantages, especially in the field of machine learning like the prediction of diseases. The COVID-19 epidemic is still having an impact on every facet of life and necessitates a fast and accurate diagnosis. Early detection of COVID-19 is exceptionally critical to saving the lives of human beings. The need for an effective, rapid, and precise way to reduce consultants' workload in diagnosing suspected cases has emerged. This paper presents a proposed model that aims to design and implement an automated model to predict COVID-19 with high accuracy in the early stages. The dataset used in this study considers an imbalanced dataset and converted to a balanced one using Synthetic Minority Over Sampling Technique (SMOTE). Filter-based feature selection method and many machine learning algorithms such as K-Nearest Neighbor, Support Vector Machine, Decision Tree, Logistic Regression, and Random Forest (RF) is used in this model. Since the best classification result was achieved by using the RF algorithm, and this algorithm was optimized by tuning the hyperparameters. The optimized RF enhanced the accuracy from 98.0 to 99.5.

Keywords

• COVID-19 SMOTE
• tuning hyperparameters
• Filter-based feature selection
• Random Forest
• Logistic Regression
• Decision tree