摘要
时间序列的相似性度量是时间序列数据挖掘研究中的一个重要问题,是进行序列查询、分类、预测的一项基础工作,寻求一种好的度量对提高挖掘任务的效率和准确性有着至关重要的意义.文章提出了基于关键点分段的KT分段线性模式表示和基于时间序列模式表示的KT动态模式匹配距离,采用1NN分类方法,设计了子序列查询实验,对欧氏距离、动态弯曲距离和基于KT模式的动态匹配距离进行了准确率和误报率的分析比较,结果显示该度量方法具有更高的准确性.
Similarity measure of time series is an important problem in data mining.It is basic for sequence query,classification and prediction.An appropriate measure is vital to improve the accuracy and efficiency of mining tasks.According to the important segment KT representation,a new similarity measure,called KT Dynamic Pattern Matching distance（KT-DTW）,is proposed.In 1NN classification,the study respectively compares Euclidean distance,DTW and KT-DTW to an accurate rate and false alarm rate.The results showed that the measurement method has higher accuracy.
出处
《宁德师范学院学报（自然科学版）》
2011年第4期344-347,共4页
Journal of Ningde Normal University(Natural Science)
关键词
时间序列
相似性度量
分类
time series
similarity measure
classification
