摘要
针对复杂系统产生的时间序列,研究其局部关联特征比研究系统全局特征模型具有明显的优势.为研究时间序列内部或局部形态的关联特征,首先借助FCM来软化时间序列属性论域的划分边界,然后,采用改进的布尔型属性关联规则并行挖掘算法来发现频繁模糊属性集,最后由多个处理器并行地产生满足最小模糊信任度的模糊关联规则.提出了基于FCM聚类的时间序列模糊关联规则的并行挖掘算法,并通过实验验证了算法的有效性.
On the occasion of dealing with time series from complex system,the investigation of series＇local patterns and local relationship has distinct superiority over traditional global models.In order to find rules relating to inside or local patterns in a time series,fuzzy C-means（FCM）clustering is used to soften the effect of sharp boundaries of delegate of each local sub-series.Then,the parallel algorithm for mining Boolean association rules is improved to discover frequent fuzzy attributes set.Finally,the fuzzy association rules with least fuzzy confidence are parallelly generated by all processors.The practical calculation results show that the mining of fuzzy association rules from time series based on FCM clustering is effective.
出处
《大连理工大学学报》
EI
CAS
CSCD
北大核心
2010年第5期806-810,共5页
Journal of Dalian University of Technology
基金
国家自然科学基金资助项目(10771092)
"九七三"国家重点基础研究发展规划资助项目(2004CB318000)
教育部专项研究课题资助项目(2007110)
关键词
数据挖掘
时间序列
模糊关联规则
并行
data mining
time series
fuzzy association rules
parallel
