演化异质信息网络集成关键技术研究

演化异质信息网络是带有时间约束的包含多种类型实体或关系的图模型，被广泛应用于建模演化数据对象及其动态的关系。演化异质信息网络的集成，能够提高异质信息管理与分析的全面性与精准性，具有重要意义。演化异质信息网络具有源数据低质性、语义多变性、结构动态性等特点，这些独有的特点导致集成演化异质信息网络面临全新的挑战：包括演化异常值与缺失值的合理判定、本体演化元路径抽取与对齐、演化实体对齐、演化事件融合、演化信息属性融合等方面。现有相关的静态集成技术与图数据集成技术均无法满足演化异质信息网络集成的需求，甚至存在研究空白。因此，本项目围绕演化异质信息网络集成展开研究，包括演化数据清洗，演化元路径的定义、抽取与对齐，演化实体对齐，演化关系与演化属性融合等理论和技术的研究；设计与实现原型系统并评测性能。本项目的研究将对演化异质信息网络的集成问题产生重要的理论意义与实际应用价值。

Evolving heterogeneous information network (HIN) is a type of time-constrained graph model, which contains multiple types of entities and relationships. Evolving HINs are widely applied to model evolving data objects with dynamic relationships. The integration of evolving heterogeneous information networks improves the comprehensiveness and accuracy of heterogeneous information management and analysis, which is of great significance. The characteristics of evolving HINs include low quality of source data, semantic variability, and dynamic structures. Thus, integrating evolving HINs faces brand new challenges including the determination of evolving outliers and missing values, the extraction and alignment of evolving meta-path in the ontology layer, the alignment of evolving entity, the fusion of evolving events and attributes. Existing integration methods of static data and graph data cannot meet the needs of evolving HIN integration. Therefore, this project focuses on the research on the integration of evolving HIN, including the cleaning of evolving data, the definition, extraction and alignment of evolving meta-paths, the alignment of evolving entities, the fusion of evolving relationships and attributes. Besides, we design and realize a prototype system of evolving HIN integration. The research of this project will have important theoretical significance and practical application value for the integration of evolving HINs.