{"title":"Tuck-KGC: based on tensor decomposition for diabetes knowledge graph completion model integrating Chinese and Western medicine.","authors":"Jiangtao ZhangSun, Yu Xin Yang, Beiji Zou, Qinghua Peng, Xiao Xia Xiao","doi":"10.7717/peerj-cs.2522","DOIUrl":null,"url":null,"abstract":"<p><p>The medical knowledge graph is essential for intelligent medical services, encompassing personalized diagnostics, precision therapies, and intelligent consultations, among others. However, medical knowledge graphs frequently suffer from incompleteness, primarily due to the absence of certain entities or relationships. The incomplete nature of knowledge graphs poses substantial challenges to these tasks. Knowledge graph completion technology is instrumental in addressing this issue. Specifically, tensor decomposition-based approaches for knowledge graph completion embed entities and relationships into the vector space, where tensor decomposition computations are employed to predict missing relationships within the knowledge graph. However, the tensor representation of entities and their relationships often overlooks crucial entity type information, potentially resulting in an abundance of irrational relationships during the prediction process. To mitigate this, we propose the Tucker Decomposition Knowledge Graph Completion (Tuck-KGC) method, which incorporates entity types into the tensor decomposition framework. This method maps the types of medical entities to vectors, which are seamlessly integrated into the knowledge graph representation learning process. This allows the model to thoroughly absorb entity information, thereby enhancing the accuracy of link prediction. To evaluate the Tuck-KGC, we built the Dia dataset, a knowledge graph tailored for precision medical analysis, which integrates both Traditional Chinese Medicine and Western medicine perspectives. The Dia dataset encompasses 10,294 entities with 214 relationships, covering a comprehensive spectrum including diseases, treatments, clinical manifestations, complications, etiology, and so on. Building upon the Dia dataset, experimental results indicate that the Tuck-KGC model boosts link prediction accuracy by roughly 8%, affirming the efficacy of incorporating entity type information into the model.</p>","PeriodicalId":54224,"journal":{"name":"PeerJ Computer Science","volume":"11 ","pages":"e2522"},"PeriodicalIF":3.5000,"publicationDate":"2025-02-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11888843/pdf/","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"PeerJ Computer Science","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.7717/peerj-cs.2522","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2025/1/1 0:00:00","PubModel":"eCollection","JCR":"Q2","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
引用次数: 0
Abstract
The medical knowledge graph is essential for intelligent medical services, encompassing personalized diagnostics, precision therapies, and intelligent consultations, among others. However, medical knowledge graphs frequently suffer from incompleteness, primarily due to the absence of certain entities or relationships. The incomplete nature of knowledge graphs poses substantial challenges to these tasks. Knowledge graph completion technology is instrumental in addressing this issue. Specifically, tensor decomposition-based approaches for knowledge graph completion embed entities and relationships into the vector space, where tensor decomposition computations are employed to predict missing relationships within the knowledge graph. However, the tensor representation of entities and their relationships often overlooks crucial entity type information, potentially resulting in an abundance of irrational relationships during the prediction process. To mitigate this, we propose the Tucker Decomposition Knowledge Graph Completion (Tuck-KGC) method, which incorporates entity types into the tensor decomposition framework. This method maps the types of medical entities to vectors, which are seamlessly integrated into the knowledge graph representation learning process. This allows the model to thoroughly absorb entity information, thereby enhancing the accuracy of link prediction. To evaluate the Tuck-KGC, we built the Dia dataset, a knowledge graph tailored for precision medical analysis, which integrates both Traditional Chinese Medicine and Western medicine perspectives. The Dia dataset encompasses 10,294 entities with 214 relationships, covering a comprehensive spectrum including diseases, treatments, clinical manifestations, complications, etiology, and so on. Building upon the Dia dataset, experimental results indicate that the Tuck-KGC model boosts link prediction accuracy by roughly 8%, affirming the efficacy of incorporating entity type information into the model.
期刊介绍:
PeerJ Computer Science is the new open access journal covering all subject areas in computer science, with the backing of a prestigious advisory board and more than 300 academic editors.