时间序列度量的斜率偏离距离方法研究  

Study of distance-measuring method in time series based on slope departure

作　　者：梁建海[1] 张建业[2] 杨峰[1] 潘泉[1] 

机构地区：[1]西北工业大学自动化学院,西安710072 [2]空军工程大学,西安710051

出　　处：《计算机工程与应用》2009年第22期4-6,共3页Computer Engineering and Applications

基　　金：国家自然科学基金No.60634030;No.60702066;国家部委预研基金资助~~

摘　　要：针对数据挖掘领域中时间序列的相似性度量问题,提出一种斜率复合偏离距离方法。以大量噪声的高维多元时间序列数据为目标,提出了一种基于斜率偏离度的时间序列相似性度量方法。该方法主要是在分段线性的基础上,基于角度和斜率进行偏离度计算,解决普通斜率距离度量的局限性,物理意义更为明确,实际度量更为准确。证明了斜率复合偏离的完备性和连续性,最后用仿真算例对算法的有效性进行了验证。Aiming at the measuring method of similarity search in time series,a similarity measuring method based on slope departure is proposed in this paper.To dispose the time series data with high dimension,multi-property,noise and default,a similarity measuring method is used to caculate the slope departure based on PLR solving the localization of slope distance measurement.In fact,the physical concept of this method proves more clearly and nicely.The maturity and the continuity of slope departure are proved.The example shows its validity.

关 键 词：时间序列 斜率偏离 距离度量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]