Arbovirose em cidade média do Ceará: análise climática e aplicabilidade de redes neurais artificiais para previsão de casos de dengue em Sobral-CE

Abstract

As it happens in big cities, the growth of medium-sized towns has been followed by socioenvironmental problems. In Ceara state, there are seven towns which have socioconomic importance and influence in their regions classified as medium-sized towns that have also presented some environmental and public health challenges.Therefore, the increase of arbovirus in Brazilian cities is an obstacle that can be related to climate and socioeconomic factors. The research is inserted in the Health Geography perspective and aims to identify and analyse the most substantial climate variables during the arbovirus occurrences in Sobral, a medium-sized town in Ceara. To this end, it wiil be applied the methodology of Socioenvironmental Geography by Medonça (2001), the quantile technique and the use of Artificial Neural Networks (Random Forest, LSTM, and Ensemble) were used to efficiently support public health policies for the local population. It was found that 75% of above-average dengue cases were associated with years with normal rainfall, and 65% were years with La Niña activity. The most populous and economically dynamic neighborhoods were those with the most dengue cases in Sobral. Regarding predictive modeling, the LSTM did not perform well throughout the testing and evaluation stages; the Ensemble model performed intermediately. However, the Random Forest model, after two adjustments, achieved a coefficient of determination that predicted dengue cases with 80% accuracy. These results contribute significantly as a reference for the literature and public health initiatives, especially in cities with more limited climatic and socioeconomic information