Seif Eddine Akermi - MSc, L’IFORCE, Faculty of Mathematics, University of Sciences and Technology Houari Boumediene, Algiers, Algeria
Mohamed L’Hadj - MD, L’IFORCE, Faculty of Mathematics, University of Sciences and Technology Houari Boumediene, Algiers, Algeria
Schehrazad Selmane - PhD, Beni Messous University Hospital Centre, Ministry of Health, Population and Hospital Reform, Algiers, Algeria

Background: Brucellosis runs rampant endemically with sporadic outbreaks in Algeria. The present studyaimed to provide insights into the epidemiology of brucellosis and compare the performance of someprediction models using surveillance data from Tebessa province, Algeria.Study Design: A retrospective study.Methods: Seasonal autoregressive integrated moving average (SARIMA), neural network autoregressive(NNAR), and hybrid SARIMA-NNAR models were developed to predict monthly brucellosis notifications.The prediction performance of these models was compared using root mean square error (RMSE), meanabsolute error (MAE), and mean absolute percentage error (MAPE).Results: Overall, ۱۳ ۶۷۰ human brucellosis cases were notified in Tebessa province from ۲۰۰۰-۲۰۲۰ with amale-to-female ratio of ۱.۳. The most affected age group was ۱۵-۴۴ years (۵۶.۲%). The cases were reportedthroughout the year with manifest seasonality. The annual notification rate ranged from ۳۰.۹ (۲۰۱۳) to۲۴۶.۷ (۲۰۰۵) per ۱۰۰ ۰۰۰ inhabitants. The disease was not evenly distributed, rather spatial and temporalvariability was observed. The SARIMA (۲,۱,۳) (۱,۱,۱)۱۲, NNAR (۱۲,۱,۶)۱۲, and SARIMA (۲,۰,۲) (۱,۱,۰)۱۲-NNAR (۵,۱,۴)۱۲ were selected as the best-fitting models. The RMSE, MAE, and MAPE of the SARIMA andSARIMA-NNAR models were by far lower than those of the NNAR model. Moreover, the SARIMA-NNNARhybrid model achieved a slightly better prediction accuracy for ۲۰۲۰ than the SARIMA model.Conclusion: As evidenced by the obtained results, both SARIMA and hybrid SARIMA-NNAR modelsare suitable to predict human brucellosis cases with high accuracy. Reasonable predictions, along withmapping brucellosis incidence, could be of great help to veterinary and health policymakers in thedevelopment of informed, effective, and targeted policies, as well as timely interventions.

Human brucellosis, Neural network auto-regressive, model, Prediction, SARIMA model, Tebessa province