交叉熵相似性度量在水文时间序列匹配中的应用  

Application of Similarity Measure of Cross Entropy in Hydrological Time Series

作　　者：伏潜[1] 朱跃龙[1] 万定生[1] 

机构地区：[1]河海大学计算机与信息学院,南京210098

出　　处：《计算机与数字工程》2015年第2期164-167,共4页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:51079040);水利部948项目(编号:201016)资助

摘　　要：基于交叉熵的相似度查询算法利用小波变换对数据进行预处理,通过PAA算法对数据进行降维,求得降维后的水位数据的近似概率分布情况。通过交叉熵距离来度量两个序列间的概率分布差异。以鄱阳湖星子站日水位数据为例,验证了基于交叉熵的相似度查询算法在降低时间复杂度的前提下较准确地找出给定时间段相似的水文过程。Similarity search algorithm based on cross entropy uses wavelet transform for data preprocessing and use PAA to reduce dimensionality .The approximate probability distribution of the water level data is obtained .By cross‐entropy distance ,the probability distribution of the difference between two sequences are measured .Taking Kang Hill station of Poy‐ang Lake water level data as an example ,this algorithm verifies the cross‐entropy‐based similarity search algorithm to accu‐rately identify a given period of time compared to a similar hydrological processes under the premise of reducing the time com‐plexity .

关 键 词：时间序列 交叉熵 相似性度量 小波变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]