DeepO: An ontology-based deep learning system for disease prediction

Abstract

Recently, there is a lot of research about building supporting diagnosis and treatment in medical care using deep learning. In this paper, we propose a method for building a deep learning model base on ontology structure for disease prediction (DeepO). The ontology structure will be automatically built based on the improvements of formal concepts analysis and pattern structure. We use deep restricted Boltzmann machine (DRBM) for the disease prediction task in DeepO. By using OCA in DeepO, our model learns more knowledge than RBM model. In experiments, we build a digestive disease diagnosis model with Vietnamese patient dataset.