Abstract: To solve the problem of scattered and heterogeneous data integration and fusion in the process of enterprise data space construction,a method of entity association relationship construction based on dynamic probe was proposed.By deploying probe points between business logic layer and data access layer of different application systems,key information such as global data structure,access log and related data was collected dynamically.Entity association model oriented to enterprise data space was constructed from three levels of pattern,instance and log.In the pattern association level,the multi-dimensional similarity discrimination algorithm combined with semantic analysis was used to realize the rapid fusion of similar entities;in the case association level,for the structured data such as numerical type,character type and the unstructured data such as long text type,the method based on feature vector similarity analysis and deep learning was used to complete the association matching of different entities;in the log association level,the association between different entities and attributes was established by analyzing the equivalence relations in the data access log.In addition,to solve the problem of uncertainty in the process of association construction,the final entity association construction scheme was given by using the reasoning model based on fuzzy logic.

Key words: entity association, data space, fuzzy logic, dynamic probe

摘要： 为解决企业数据空间构建过程中分散、异构的数据集成与融合问题,提出一种基于动态探针的实体关联关系构建方法。通过在不同应用系统的业务逻辑层和数据访问层之间部署探测点,动态收集全局数据结构、相关数据和访问日志等关键信息,分别从模式、实例和日志3个层面构建面向企业数据空间的实体关联模型。在模式关联层面,采用结合语义分析的多维相似度判别算法实现相似实体的快速融合;在实例关联层面,针对数值型、字符型等结构化数据与长文本型等非结构化数据,利用基于特征向量相似度分析与深度学习相结合的方法完成对不同实体的关联匹配;在日志关联层面,通过分析数据访问日志中的等价关系建立不同实体和属性之间的关联。此外,为解决关联关系构建过程中的不确定性问题,采用基于模糊逻辑的推理模型,给出最终的实体关联构建方案。

关键词: 实体关联, 数据空间, 模糊逻辑, 动态探针