III: Medium: Collaborative Research: KMELIN: Knowledge Mining and Embedding Learning for Complex Dynamic Information Networks

复杂动态信息网络（CDINs）由与各种依赖关系高度相关的数据对象组成，例如患者-医生交互或患者-药物-保险索赔。作为CDIN节点的每个数据对象都具有丰富的内容，例如患者的生物特征信息、疾病症状或医院后勤。数据对象及其关系也在不断发展和变化。许多健康、社会、物理和生物系统都共享CDIN的本质，即单个节点的多面性和动态性对整个复杂和不断发展的网络建模提出了重大挑战。虽然数据关系比以往任何时候都更加丰富和全面，但现有系统大多是关系数据库驱动的，无法整合复杂的网络数据关系进行大数据分析。本项目旨在为CDIN设计一个知识挖掘和嵌入学习平台，该平台将(1)提取和表示健康领域中结构复杂、内容丰富的CDIN信息；(2)在CDIN网络上进行知识挖掘，包括聚类和分类；(3)利用cdin实现特征嵌入学习，使用户可以与cdin交互访问内容；(4)为医院再入院决策支持提供原型系统。该项目的方法范围不仅将丰富挖掘复杂结构和丰富内容网络的算法和解决方案，而不是静态网络，而且还将使现有的卫生信息系统从传统数据库转向以网络为中心的系统。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Complex dynamic information networks (CDINs) consist of data objects that are highly correlated with a variety of dependency relationships, such as patient-physician interactions or patient-medication-insurance claims. Each data object as a CDIN node has rich contents, such as biometric information of a patient, disease symptoms, or hospital logistics. Data objects and their relationships also continuously evolve and change. Many health, social, physical, and biological systems share the CDIN essence that the multifaceted and dynamic nature of individual nodes imposes significant challenges for modeling a complex and evolving network as a whole. Although data relationships are becoming rich and comprehensive than ever, existing systems are mostly relational-database driven, and cannot integrate complex relationships of networked data for Big Data analytics.This project aims to design a knowledge mining and embedding learning platform for CDINs that will (1) extract and represent complex structure and rich-content information in the health domain as a CDIN; (2) perform knowledge mining, including clustering and classification, on CDIN networks; (3) enable feature embedding learning with CDINs, so the users can interact with CDINs for content access, and (4) provide a prototype system for hospital re-admission decision support. The spectrum of the methods from the project will not only enrich algorithms and solutions for mining complex structure and rich content networks, as opposed to static networks, but also shift existing health information systems from traditional databases towards becoming network centered systems.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.