Résumé
Malaria elimination, one of the Sustainable Development Goals of the United Nation, is challenged by cross-border context specificities. At the French Guiana-Brazil border, a system was developed to harmonized epidemiological data providing by the two countries. This study evaluates the feasibility of using such harmonized data to build a cross-border early warning system. To this end, the study compared ARI-MAX and LSTM approaches. Time-lagged meteorological data were introduced to improve the forecasts. LSTM outperformed ARIMAX, with a 10 to 39% decrease of Mean Absolute and Root Means Square Errors, and better predicted low case numbers. Meteorological data improved significantly model predictions, by considering time-lags from 3 to 7 weeks compatible with the knowledge found in the literature. This study demonstrated the feasibility of building a cross-border malaria early warning system, that would significantly contribute to malaria control and elimination.