Research on Predicting Food Allergy Based on Recurrent Neural Network

Abstract

Food allergy is hard to detect because it is varying among individuals. Traditional methods based on clinical history or symptom monitoring and diagnosis are often not sensitive. Therefore, it is necessary to study new and objective methods to predict food allergy. The hygiene hypothesis proposes that early exposure and exposure to the microbial environment could reduce the possibility of suffering from allergic diseases. Therefore, exploring the microbiome-based prediction method is expected to make up for the shortcomings of traditional methods and provide effective information for early intervention. In response to the above problems, we propose a recurrent neural network to analyze microbiome time-series data. Experimental results show that RNNs are significantly better than traditional machine learning methods. In addition, we analyze the impact of different feature selection methods on classification and introduce a specific method to determine the dimension of important features using autoencoder.