摘要
针对多源异构数据集成易受数据随机性和不确定影响的问题,提出了基于多尺度时空聚类的动态传感信息多源异构数据集成方法。在可编程阵列逻辑下,构建多源异构数据采集框架,实现数据采集接口与帧格式的实时可配置。通过机器学习方法分析数据,依据对象之间类似程度划分数据集,分析任意2个对象之间的相异性,构建相异同数据结构矩阵。通过时间反转处理方式重构高维空间,实现多源异构数据时空结构的加权处理。由实验结果可知,与高效实体识别、规则化有限学习机(Regularized finite learning machine,RELM)方法相比,所提方法数据集成效果较好,且数据集成时间较短,最短为25s。实验结果验证了所提方法具有高效集成效果,具备较好的应用价值。
A dynamic sensing information multi-source heterogeneous data integration method based on multi-scale spatiotemporal clustering is proposed to address the issue of susceptibility to data randomness and uncertainty in multi-source heterogeneous data integration.Build a multi-source heterogeneous data collection framework under programmable array logic to achieve real-time configurable data collection interfaces and frame formats.Analyze data through machine learning methods,divide the dataset based on the similarity between objects,analyze the dissimilarity between any two objects,and construct a dissimilarity and similarity data structure matrix.Reconstruct high-dimensional space through time reversal processing to achieve weighted processing of the spatiotemporal structure of multi-source heterogeneous data.According to the experimental results,compared with efficient entity recognition and RELM methods,the proposed method has better data integration performance and shorter data integration time,with a minimum of 25 seconds.The experimental results verify that the proposed method has efficient integration effect and good application value.
作者
谢斌生
郭成根
Xie Binsheng;Guo Chenggen(Fujian Polytechnic of information Technology,FuJian FuZhou 350003;Fujian Jinjiang Huaqiao Vocational School,Fujian Jinjiang 362200)
出处
《现代科学仪器》
2023年第6期211-216,共6页
Modern Scientific Instruments
关键词
多尺度时空聚类
动态传感信息
多源异构
数据集成
multi-scale spatiotemporal clustering
Dynamic sensing information
Multi-source isomerism
data Integration
