海洋水文观测数据聚类  

CLUSTERING OF MARINE HYDROLOGICAL OBSERVATION DATA

作　　者：闫可 程文芳[2] 

机构地区：[1]复旦大学计算机科学技术学院,上海201203 [2]中国极地研究中心,上海200136

出　　处：《计算机应用与软件》2017年第11期39-43,90,共6页Computer Applications and Software

基　　金：极地海洋环境监测网系统研发及应用示范项目(201405031)

摘　　要：在科学考察中,数据的获取受自然环境因素以及监测成本影响较大,实际布放的监测点的数量和位置可能无法达到预期,并且所采集的数据集中通常包含了多种监测要素,利用数据分析来弥补因自然环境影响而造成的数据缺失并找出数据变化规律显得尤为重要。以南极普里兹湾水文数据为研究对象,利用空间插值的方法,来弥补数据不足和监测点稀疏的问题,再将改进的动态时间弯曲距离算法用于具有多要素特性的水文深度序列相似度衡量,实验结果表明相较于传统的欧氏距离相似度衡量更为准确。基于所提出的相似度衡量算法,对普里兹湾水文数据进行聚类,并获得了每个簇的空间分布情况。In the course of scientific investigation, the acquisition of data is greatly affected by the natural environment factors and the monitoring cost. The actual number and location of monitoring points may not be able to meet the expectations and the collected data set usually contains a variety of monitoring elements. It is particularly important to use data analysis to compensate for lack of data caused by the natural environment and find out the law of data change. Based on the hydrological data of Prydz Bay in Antarctica, the use of spatial interpolation method to make up lack of data and sparse monitoring points, then the improved Dynamic time warping distance algorithm is applied to the similarity measure of hydrological depth series with multi-element. The experimental results show that similarity measurement algorithm is more accurate than the traditional Euclidean distance. Based on the similarity measurement proposed in this paper, the Prydz Bay hydrological data are clustered and the spatial distribution of each cluster is obtained.

关 键 词：水文数据 空间插值 动态时间弯曲 相似度衡量 深度序列 K-MEANS 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]