A Hybrid Time Series–Regression Model for Tuberculosis Forecasting in Resource-Limited Settings

Tuberculosis (TB) is still a serious public health issue in Sudan, especially in Gedaref State, because of limited medical facilities and inadequate disease reporting. This experiment develops a forecasting model by employing Seasonal Trend decomposition using LOESS (STL) and linear regression in combination, relying on the weekly tests to improve TB prediction. The model improves the accuracy of its forecasts by combining time series information with the details of the daily operations of the health system. Weekly data from Gedaref showed that the STL + regression approach performed better than ARIMA, reducing the root mean squared error (RMSE) from 2986.85 to 540.95, an improvement of about 81.9%. The model also remained flexible to fluctuations in testing volume. The findings illustrated that hybrid statistical methods have been proved to be reliable and practical in forecasting TB cases in situations where limited resources exist, providing a strong base for overseeing TB and other communicable diseases.