Deep Neural Network Based on Translation Model for Diabetes Knowledge Graph

Knowledge base completion (KBC) can predict new facts according to existed facts in knowledge base. However, most work in KBC is limited to simple neural networks and usually focuses on commonsense knowledge base where data resources are from public websites. Medical records related to diabetes from...

Full description

Saved in:
Bibliographic Details
Published in2017 Fifth International Conference on Advanced Cloud and Big Data (CBD) pp. 318 - 323
Main Authors Suna Yin, Dehua Chen, Jiajin Le
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2017
Subjects
Deep Neural Network
Diabetes
Diabetes Knowledge Graph
Diseases
Hospitals
Knowledge Base Completion
Knowledge based systems
Knowledge engineering
Medical diagnostic imaging
Neural networks
Translation Model
Online AccessGet full text

Cover

More Information
Summary:Knowledge base completion (KBC) can predict new facts according to existed facts in knowledge base. However, most work in KBC is limited to simple neural networks and usually focuses on commonsense knowledge base where data resources are from public websites. Medical records related to diabetes from hospital showed potential advantages in both quality and quantity, which is significant to be mined for preventing or delaying diabetes and its complications. Thus, this paper designs and constructs Diabetes Knowledge Graph from electronic medical records in Shanghai Ruijing Hospital, and proposes a deep neural network based on translation model for completion in the diabetes knowledge graph. We manually evaluate our trained model's ability in accuracy, recall and F1-scores, finding that it has a good stability and performance in diabetes knowledge base completion.
DOI:10.1109/CBD.2017.62