Neural Network Multi-label Learning Based on Enhancing Pairwise Labels Discrimination for Obstetric Auxiliary Diagnosis

Abstract

The data-driven medical health information processing has become a new development direction, especially the auxiliary diagnosis based on the electronic medical records (EMRs), which is of great significance to improve population health. In this paper, to obtain excellent obstetric auxiliary diagnostic results, the Chinese obstetric EMRs is analyzed and processed, and finally the auxiliary diagnosis task is transformed into a multi-label classification problem. Moreover, two effective global error functions are proposed by enhancing pairwise labels discrimination to improve the Backpropagation for Multi-label Learning (BP-MLL) that depends on the neural network model. The experiment results of some public multi-label datasets and the Chinese obstetric dataset show that the two error functions have better overall performance compared with BP-MLL original error function and some well-established multi-label learning algorithms.