Predicting Dengue Fever in Brazilian Cities

Dengue Fever is an increasingly serious public health concern both in Brazil and globally. In the absence of a universal vaccine or specific treatments, prevention relies on vector control and disease surveillance. Accurate and early forecasts can help reduce the spread of the disease. In this study, we develop a model to predict the number of Dengue Fever cases in Brazilian cities one month ahead. We compare different machine learning approaches as well as different sets of input features based on epidemiological and meteorological data. We find that different models work best in different cities, and a random forests model trained on data of historical Dengue cases performs best overall. It produces lower aggregate errors than a seasonal naive baseline model, Gradient Boosting Regression, feed-forward Neural Networks, and Support Vector Regression. Predictions on an unseen test set are on average within 11.5 cases for the median city. Mean absolute errors on the hold-out test set are reduced to 10.8 for the median city when selecting the optimal combination of algorithm and input features for each city individually.