A set of independent variables for time series regression tasks of pandemic scenarios based on Covid-19

Abstract

We conduct research on a wide range of variables encompassing factors, governmental interventions, and events, with potential implications on the dissemination dynamics of SARS-CoV-2, either positively or negatively impacting, based on the trend of cases and deaths, that can be used as features in reducing errors in regression tasks under pandemic contexts such as COVID-19. The study employs a machine learning framework to discern pertinent characteristics and parameters instrumental in elucidating the intricate relationship between COVID-19 incidence and mortality rates across diverse spatial and temporal resolutions. The strategy involves conducting a systematic feature survey for data collection and cleaning, clustering, reducing data dimensionality, applying machine learning techniques with the support of models such as Seasonal ARIMA and LSTM, and evaluating model generalization through component and network analysis. Using a specific set of independent variables to predict COVID-19 deaths can improve accuracy, reduce standard deviation from actual values, and prevent sub-specification of the problem. The main contributions of this work entail a comprehensive investigation into the causative factors and potential interrelations among phenomena, governmental interventions, and events influencing the propagation dynamics of the virus within urban locales, spanning cities and countries. By furnishing complementary resources, this research offers valuable support to governmental bodies and authoritative entities in navigating decision-making processes amidst potential pandemic scenarios, such as the COVID-19 crisis, which, while still present, has a lower mortality rate. Our findings not only offer insights into prospective predictors of COVID-19 transmission but also furnish a structured framework for enhancing the efficacy of time series regression models within the complex landscape of pandemic scenarios.