摘要
传感器技术发展促进各行各业产生了大量多源数据,且这些数据还在不断发生变化。当多源数据(分布信息系统)增加了一些属性后,传统约简算法需要重复计算数据且不能有效实现多源数据融合,导致计算动态多源数据约简花费时间较多,计算效率不高。为了克服传统约简算法的缺陷,设计了基于多源数据矩阵增量约简算法。介绍了一些分布信息系统的相关理论知识,给出了多源数据等价关系矩阵融合的计算方法。当多源数据增加了一些属性后,讨论了动态多源数据增量机制、融合方法及矩阵增量约简算法。分别利用矩阵增量和矩阵非增量约简方法对4个UCI数据集进行测试,测试结果验证了所提出的矩阵增量方法能够快速解决动态多源数据约简更新问题。
Since multi-resource data processing has been involved in many scientific research fields,the traditional attri-bute reduction algorithm often needs to run from scratch when some attributes are added into the multi-resource data and thus it consumes a lot of computational time.In response to the defect,a matrix-based incremental attribute reduction algo-rithm is proposed when multiple attributes are added into the multi-resource data.This paper introduces some definitions and conceptions of distributed information system,and data fusion method for matrix of equivalence relation of multi resource data is proposed.The incremental mechanisms and data fusion techniques for multi-resource data and mathematical expression are given and the corresponding incremental attribute reduction algorithm is proposed when some attributes are added into the multi-resource data.This paper compares the computation time between the non-incremental attribute reduction algorithm and the incremental attribute reduction algorithm on the 4 data sets from UCI and the experimental results show that the incremental attribute reduction algorithm can deal with attribute reduction of dynamic multi-resource data efficiently.
作者
徐岩柏
景运革
XU Yanbai;JING Yunge(School of Maths&Information Technology,Yuncheng University,Yuncheng,Shanxi 044000,China)
出处
《计算机工程与应用》
CSCD
北大核心
2022年第3期195-200,共6页
Computer Engineering and Applications
基金
国家自然科学基金(61703363)
山西省应用基础研究计划(201801D121148)
运城学院院级项目(YQ-2017028)。
关键词
多源数据
增量学习
属性约简
关系矩阵
multi-resource data
incremental learning
attribute reduction
relation matrix
