摘要
针对动态多元时间序列(MTS)上的状态转移模式(STAP)挖掘问题,提出一种准确高效的三支增量挖掘算法(3IU-STAP)。该算法根据原始数据、已有频繁STAP和增量数据构造出候选模式,使用频繁阈值将其划分至正域、负域以及边界域。仅边界域中的候选模式需要延迟决策,即通过扫描数据集来判断其是否频繁。准确性方面,设计了增量数据补齐技术,获得候选模式实际出现次数。效率方面,使用了向下封闭性质来控制候选模式数量,尽可能减少对数据的扫描。在4个真实数据集上的实验结果表明,与非增量方法相比,3IU-STAP可以得到准确结果,同时显著提高效率。
An accurate and efficient three-way decision based incremental mining algorithm for mining state transition patterns(STAP)was proposed for dynamic MTS.Based on the original data,existing frequent STAPs and incremental data,candidate STAPs were constructed and divided into positive,negative and boundary regions through a frequent threshold.Only those STAPs in boundary region was delayed in decision making.The data sets were scanned to check whether they were frequent.In terms of accuracy,incremental data completion technique was designed to obtain the actual occurrences of candidate STAPs.In terms of efficiency,the downward closure property was introduced to control the number of candidate patterns and minimize the scanning of data.Results of experiments undertaken on four real data sets showed that 3IU-STAP could achieve accurate results while significantly improving time performance compared with non-incremental methods.
作者
曾圣超
张智恒
闵帆
张紫茵
ZENG Shengchao;ZHANG Zhiheng;MIN Fan;ZHANG Ziyin(Department of Computer Science,Southwest Petroleum University,Chengdu 610500,China;School of Information and Engineering,Sichuan Tourism University,Chengdu 610100,China)
出处
《郑州大学学报（理学版）》
CAS
北大核心
2020年第1期16-23,共8页
Journal of Zhengzhou University:Natural Science Edition
基金
国家自然科学基金项目(41604114)
四川省科技创新团队项目(2019JDTD0017)
四川省科技厅项目(2019YJ0314)
关键词
多元时序
三支决策
序列模式
增量挖掘
multivariate time-series
three-way decision
sequence pattern
incremental discovery
