COVID-19 Prediction and Detection Using Machine Learning Algorithms: Catboost and Linear Regression
Justine Shinjae Kim
Abstract
A global pandemic COVID-19 has been rapidly spreading, and the predictions for infected rate shows how the cases will increase or decrease. Even though the number of people who get the corona vaccine is increasing, COVID-19 has been a serious worldwide problem. As machine learning and deep learning were implemented to predict COVID-19 in recent days, machine learning to predict the number of confirmed and death cases of COVID-19 was used. Prediction graphs of our proposed model play a crucial role for preventing more people getting infected. The project collected the number of daily infected cases in New York from March 21th 2020 to March 6th 2021. For precise results, the dataset in 6 different kinds of the machine learning methods was used. The methods were Decision Tree, Random Forest, Linear Regression, Gradient Boosting, XGboosting, and LGBM. RMSE and MAE values fluctuated from 9.95 to 68.85 and 5.99 to 58.76. The most accurate model was Linear Regression, RMSE and MAE with 9.96 and 5.99 for death cases and 597.61 and 346.04 for confirmed cases. Therefore, those prediction graph almost matched the same as the real number graph that the project drew with an actual dataset. The other dataset was about common COVID-19 symptoms, and the Catboost model listed from the most influential factor, breathing problem. Collecting data from other areas and specifying the patients’ features could have improved the quality of the research, though overall the result was successful.