Modeling Vector Populations with Deep Learning
Each year, vector borne diseases impact hundreds of millions of people globally. Aedes aegypti is one of the most potent vectors, with the potential to transmit dengue, Chikungunya, Zika, and other arboviral diseases. Accurate forecasting of Aedes aegypti population dynamics is a crucial step for predicting and mitigating vector-borne disease outbreaks. Currently, a stochastic, agent-based computer simulation is used to predict Aedes aegypti populations, but the simulation requires high performance computing to generate predictions for substantial spatiotemporal regions. In this work, we investigate convolutional neural networks (CNN), long short-term memory networks (LSTM), and gated recurrent units networks (GRU) models as feasible alternatives to the agent-based model. We present techniques to balance the training data and discuss the performance of the deep learning models in replicating the population dynamics predicted by the agent-based model.