MOSQLIMATE: A PLATFORM FOR COMPARING ARBOVIRAL DISEASES FORECASTING MODELS

Objectives: The Mosqlimate platform aims to enable proper statistical comparisons between arboviral diseases forecasting models trained on the same data. The platform also proposes a baseline forecast model to predict the number of cases for any Brazilian city. Methods: To ensure model comparability, the platform provides an API for downloading various datasets, ensuring all models can learn from the same data. The available datasets include case notifications (dengue, Zika, and chikungunya), climate data, mosquito surveillance data, and epidemiological parameters for all Brazilian municipalities. To compare the forecasting models, the Mosqlimate API allows users to register their models for specific arboviruses, store their predictions, and visualize them, comparing them with the predictions from other models registered on the platform. The platform supports both time series and map visualizations. Results: The Mosqlimate platform is available online (https://api.mosqlimate.org/). It features multiple datasets and registered models that can be viewed in the Dashboards section. Anyone interested in using the platform can register and use it. The Infodengue project, a Brazilian early warning alert system, already utilizes some of the models available on the platform. Conclusion: The Mosqlimate platform offers data for training models to forecast arboviruses cases, visualize predictions, and compare them with other models registered on the platform. This enables the selection of more efficient models, supporting decision-making and the implementation of control measures.