摘要
由于传统的时序相似性度量方式不满足距离三角不等式关系,影响后续的相似性搜索及关联规则的获取,在时序符号化的基础上,提出一种满足三角不等式的符号化距离度量方式。与MINDIST_PAA_SAX和Sym_PAA_SAX度量方式进行比较,其结果表明,该度量方式在异常检测和相似性查询上具有较好的优越性。实验结果表明,该方法在相似性搜索及关联规则的获取方面具有更高的可信度。
The traditional similarity measures for time series fail to meet the distance triangle inequality,which further affects the subsequent similarity search and the acquisition of association rules.Based on the symbolic time series,a symbolic distance measure which met the distance triangle inequality was proposed.Compared with MINDIST_PAA_SAX and Sym_PAA_SAX,the metric based on the triangle inequality is superior in the anomaly detection and similarity queries and the experimental results show higher reliability in the similarity search and the acquisition of association rules.
出处
《计算机工程与设计》
CSCD
北大核心
2014年第12期4191-4194,4201,共5页
Computer Engineering and Design
基金
中央高校基本科研业务费专项基金项目(2011QNB23)
国家863高技术研究发展计划基金项目(2012AA011004)
教育部博士点基金项目(20110095110010
20100095110003)
关键词
时间序列
符号化表示
距离度量
相似性搜索
关联规则
time series
symbolic representation
distance metric
similarity search
association rules