Machine learning aproach for Dengue forecasting - Comparing LSTM, Random Forest and Lasso methods

Aluno(a): 

  • Elisa Mussumeci Bianor dos Santos

Data: 

12/04/2018 - 10:00

Local: 

Sala da Congregação FGV/EMAp (5º andar)

Resumo: 

We use the Infodengue database of incidence and climate time-series, to train predictive models for the weekly number of cases of dengue in 790 cities of Brazil. To overcome limitation in the length of timeseries available to train the model, we included the time series of similar cities as predictors in the model of each city. Machine Learning models to forecasting have been used in the past years and have achieved great results. In this work we will compare three machine learning models: Random Forest, Lasso and Long-short term memory neural network.

*Texto enviado pelo aluno. 

Membros da banca: 

  • Flavio Codeço Coelho (Orientador) - FGV/EMAp
  • Rodrigo dos Santos Targino - FGV/EMAp
  • Leonardo Soares Bastos - FIOCRUZ