Biomedical Data Driven COVID-19 Prediction Using Machine Learning Approach

Abstract

AbstractThe novel Corona Virus (SARS-CoV-2) is a new strain of Corona Virus that had never affected humans before. The SARS-CoV-2 virus causes the sickness termed COVID-19. COVID-19 has stretched throughout this planet, causing a global pandemic. In this challenging situation, Machine Learning can play an important role in preventing and control COVID-19 by identifying a case, predicting future spreads, and more. It can also play a pivotal role in developing drugs and vaccines to fight the disease. In this work, a machine learning approach has been used to analyze biomedical data of COVID-19 suspected patients to predict whether those patients have positive prediction results or not. To date, there is no proper treatment of COVID-19. Even there is no proper full-proof vaccine also. But this Corona Virus (SARS-CoV-2) can spread from one human to another human without their knowledge through droplets. Minimization of close human interaction is the only possible solution to restrict the spread of the Corona Virus (SARS-CoV-2). Since December 2019, this virus has been spreading throughout the world, involving almost all the countries of the world. In this situation, the machine learning approach can save us by restricting the spread of the virus. Currently there several techniques of tests (like RT-PCR) are there to find whether a person is COVID-19 positive or not. In those tests, nasal and throat swab is collected and then the test is conducted on that. It is costly as well time taking. Moreover, there is a chance that the collector person also may get infected. An alternative machine learning-based approach is proposed in this effort, which does not require close interaction with COVID-19 suspected patients. This approach relies on some symptoms of the suspected patients. As the currently running testing techniques takes a long time to generate a result, there is a possibility that the suspected patient may spread this virus among other persons before the test result come. The proposed scheme minimizes this possibility by providing an early notification of infection before the actual test result comes. If a suspected patient is notified of a positive COVID-19 infection, all precautionary measures like quarantine and contact tracing can be started before the actual test result arrives. This minimizes the spread of the Corona Virus (SARS-CoV-2) among persons, which reduces the chances of community transmission of this virus. Some common COVID-19 symptoms like—breathing problem, fever, dry cough, sore throat, headache, fatigue is considered in this effort. Some other factors which increase the possibility of being infected by COVID-19 like abroad travel history, close contacts with COVID-19 positive patients, attending large public gatherings, visiting public exposed places, whether a family person working in public exposed places, wearing a mask, sanitization after coming from the market are also considered in this effort along with information of COVID-19 symptoms. A total of 15 factors are considered in this work. This work aims to provide support to society by predicting the possibility of COVID-19 positive before the actual test result comes without any close interaction with the suspected person.KeywordsCOVID-19Machine learning techniqueSymptomsPrediction