An evolutionary deep neural network for predicting morbidity of gastrointestinal infections by food contamination

A majority of gastrointestinal infectious diseases are caused by food contamination, and prediction of morbidity can be very useful for etiological factor controlling and medical resource utilization. However, an accurate prediction is often very difficult not only because there are various types of food and contaminants, but also because the relationship between the diseases and the contaminants is highly complex and probabilistic. In this study, we use the deep denoising autoencoder (DDAE) to model the effect of food contamination on gastrointestinal infections, and thus provide a valuable tool for morbidity prediction. For effectively training the model with high-dimensional input data, we propose an evolutionary learning algorithm based on ecogeography-based optimization (EBO) in order to avoid premature convergence. Experimental results show that our evolutionary deep learning model obtains a much higher prediction accuracy than the shallow artificial neural network (ANN) model and the DDAE with other learning algorithms on a real-world dataset.