COVID-SGIS

COVID-SGIS performs dynamic monitoring and six-day temporal forecasting of cumulative COVID-19 confirmed cases and deaths using ARIMA (AutoRegressive Integrated Moving Average) models on Brazil.io National Notification System data to support epidemiological decision-making.

Key Features:

• Real-Time Data Monitoring: Captures official notifications from Brazil's 26 states and the Federal District via the National Notification System as compiled on the Brazil.io database.
• Temporal Forecasting Models: Builds ARIMA (AutoRegressive Integrated Moving Average) models trained on historical data from March–May 2020 to forecast cumulative confirmed cases and deaths.
• Evaluation Metrics: Evaluates forecasting performance using Pearson, Spearman, and Kendall correlation indexes alongside RMSE (%).
• Forecasting Capability: Produces six-day forecasts for Brazil and each federative unit with 95% confidence intervals and worst-case and best-case scenario projections.
• Error Analysis: Reports prediction error ranges for Brazil (2.56%–6.50%) and notes forecasts outside the prediction interval deviate less than 5% from worst-case scenarios.

Scientific Applications:

• Policy Guidance: Guides social policies and the planning of direct public-health interventions using short-term forecasts.
• Regional Surveillance: Supports adaptation to different regional realities by providing forecasts for individual federative units.
• Decision Support: Supports informed epidemiological decision-making by integrating official National Notification System data compiled on Brazil.io.

Methodology:

Captures Brazil.io recordings of the National Notification System for all federative units, trains ARIMA models on March–May 2020 historical data to forecast cumulative cases and deaths, evaluates forecasts with Pearson, Spearman, Kendall and RMSE (%), and generates six-day forecasts with 95% confidence intervals and scenario projections.