Dengue Fever Forecasting in Brazil

WHO data wrangling and time series analysis

This project focuses on the analysis and forecasting of Dengue Fever cases in Brazil, utilizing data from the World Health Organization (WHO).

Data Source & Wrangling

The primary dataset for this project was obtained from the Pan American Health Organization (PAHO), specifically their Dengue Fever indicators for Brazil.

The initial phase involved minor cleaning and wrangling of this dataset for epidemiological comparison. This included adding new columns to track the change in values over time, which is crucial for understanding disease progression. Redundant or unneeded values were also dropped to streamline the dataset for analysis.

Forecasting Methodology: SARIMA

Forecasting was primarily performed using the **Seasonal Auto-Regressive Integrated Moving Average (SARIMA)** model. This model was appropriate in taking into account the seasonal, non-stationary, and with more work the exogenous nature of infectious disease. SARIMA models extend ARIMA models by incorporating seasonal components, making them suitable for data with recurring patterns like disease outbreaks and annual cycles.

Currently, this is only a SARIMA model. External factors like weather, public health interventions, and population density have not yet been included. These exogenous variables play a significant role in infectious disease dynamics, and their integration would enhance model accuracy. Future work will involve integrating these external factors to enhance the model's predictive accuracy and provide a more comprehensive understanding of dengue fever dynamics.

This forecasting was specifically conducted to predict the year 2023 data, utilizing only historical data available up to March of that year.

Data Visualizations and Predictions

The model predicts, with increasingly large confidence intervals, that Dengue Fever cases will decrease until eventually having a minor spike comparable with past seasons. Uniquely, this model predicts a historically low Dengue Fever season for 2023, with predictions hitting 0 new cases for some weeks.

Dengue Fever Case Forecast
Figure 1: Forecasted Dengue Fever Cases with Confidence Intervals. The model predicts a decrease in cases, with some weeks hitting zero, indicating a potentially historically low season for 2023.

Dengue Fever total Deaths are predicted to be somewhat comparably low overall with a similar S-curve. Notably, these confidence intervals get incredibly wide, with 2022 Dengue Fever deaths being comparable and possible for this Dengue Fever season.

Dengue Fever Death Forecast
Figure 2: Forecasted Dengue Fever Deaths with Confidence Intervals. Predictions show relatively low deaths, but wide confidence intervals suggest the possibility of similar death tolls to 2022.

Contributions

All work on this project was completed by Trystan Carruth.

Email: Trystaned@gmail.com

View the full project on GitHub