Rational drug use prediction, which is used to estimate whether the drug use is reasonable in clinical medical treatment, is of great significance in combating excessive medical treatment. The feature interaction model based on electronic medical records is widely used in the field of medical behavior prediction. However, the features recorded in electronic medical records are very complex, and medical behavior has an obvious long-tail effect. The problem of feature sparseness makes traditional prediction models based on feature interaction impossible. It works well, we propose a novel business domain-based second-order relational graph embedding neural network model (SORGE-NN), which can distinguish the scope of features and mine the implicit propagation relationship, through the residual-multi-head attention layer , and perform high-order weighted combination of features, which effectively alleviates the challenges brought by feature sparse and complex features. We conduct experiments with real datasets, and the experimental results show that our proposed SORGE-NN achieves better results than current state-of-the-art prediction models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.