Forecasting dog rabies dynamics in Tunisia using time series models: insights for early warning systems

Abstract

Introduction: Rabies is endemic in Tunisia, with a rising epidemic trend being observed over the years and especially after 2012, leading to substantial economic impacts on both animal and human populations. While temporal trends have been previously documented, time series analysis offers a powerful tool for understanding this evolution. To inform evidence-based surveillance and control strategies, this study aimed to identify the most accurate time series model for forecasting monthly dog rabies cases. Methods: A time series analysis approach was conducted to model and forecast rabies cases in dogs from January 1994 to December 2023. Several forecasting models were evaluated for their performance, including Seasonal Autoregressive Integrated Moving Average (SARIMA), Error-Trend-Seasonal (ETS), neural network nonlinear autoregression (NNAR), Prophet model and the Trigonometric Exponential Smoothing State-Space Model with Box-Cox transformation, ARMA errors, Trend and Seasonal components (TBATS). Model accuracy was evaluated using root mean squared error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). Results: The TBATS model exhibited the highest forecasting accuracy compared to the other tested models. Discussion: These results indicate that TBATS is the most reliable model for short- to medium-term rabies forecasting in Tunisia. These findings highlight the potential of advanced time series modeling in improving rabies surveillance and control strategies in Tunisia, supporting more effective resource allocation and intervention planning.