Machine learning in front of statistical methods for prediction spread SARS-CoV-2 in Colombia
This work addresses COVID-19 prediction for public health planning in Colombia, but it is incremental as it applies existing methods to new data without major innovations.
The study compared SEIR, Logistic Regression, and Polynomial Regression models to predict SARS-CoV-2 spread in Colombia over 550 days, finding that machine learning methods like Polynomial Regression had lower propagation error and reduced statistical biases. It also proposed four prevention scenarios to evaluate disease parameters.
An analytical study of the disease COVID-19 in Colombia was carried out using mathematical models such as Susceptible-Exposed-Infectious-Removed (SEIR), Logistic Regression (LR), and a machine learning method called Polynomial Regression Method. Previous analysis has been performed on the daily number of cases, deaths, infected people, and people who were exposed to the virus, all of them in a timeline of 550 days. Moreover, it has made the fitting of infection spread detailing the most efficient and optimal methods with lower propagation error and the presence of statistical biases. Finally, four different prevention scenarios were proposed to evaluate the ratio of each one of the parameters related to the disease.