贺琪,武欣怡,黄冬梅,郝增周,宋巍.多视图协同的海洋多要素环境数据关联关系分析方法[J].海洋通报,2019,(5):533-542
多视图协同的海洋多要素环境数据关联关系分析方法
Multi-view analysis method for marine multi-factor environmental data association
投稿时间:2019-03-26  修订日期:2019-05-16
DOI:10.11840/j.issn.1001-6392.2019.05.007
中文关键词:  海洋多要素环境数据  多视图协同  关联关系  多维标度算法  K-means 聚类
英文关键词:marine multi -factor environmental data  multi -view collaboration  correlation  multidimensional scaling algorithm  K-means clustering
基金项目:海洋大数据分析预报技术研发基金(2016YFC1401902);国家海洋局数字海洋科学技术重点实验室开放基金(B201801029);上海市高校特聘教授(东方学者) 项目(TP2016038);上海市科委部分地方院校能力建设项目(17050501900)
作者单位E-mail
贺琪 上海海洋大学信息学院上海201306 qihe@shou.edu.cn 
武欣怡 上海海洋大学信息学院上海201306  
黄冬梅 上海电力大学上海200090  
郝增周 自然资源部第二海洋研究所卫星海洋环境动力学国家重点实验室浙江 杭州310012 hzyx80@sio.org.cn 
宋巍 上海海洋大学信息学院上海201306  
摘要点击次数: 555
全文下载次数: 165
中文摘要:
      海洋事件离不开各要素环境数据的共同作用,获取要素之间的关联关系从而进行海洋事件的预报预测,是一个亟待解决的问题。为此,本文提出一种多视图协同的关联关系分析方法来度量海洋各要素数据间的关联关系。首先,在传统平行坐标技术的基础上增加刷技术、轴排序等功能对海洋多要素数据进行初步探索,同时引入散点矩阵图展示各要素的分布;其次,以平行坐标中数据线间的角度、面积以及散点图中要素分布的距离为差异度量方式,对计算得到的差异构建相似性矩阵;再次,采用多维标度法得到原始多要素数据在低维空间中的表达;最后,使用K-means 算法对降维后的低维度数据进行聚类分析。本文提出的方法从视觉角度对数据进行分析和特征挖掘,并得到高维数据在低维空间上的可视化展示,实现了有效量化海洋数据不同要素间的相关关系。
英文摘要:
      Marine phenomenon cannot be separated from the joint action of various environmental elements. The correlation relationship between elements needs to be addressed so as to predict marine events. In this study, a multi-view collaborative correlation analysis method is proposed to measure the correlation among ocean data. Firstly, on the basis of the traditional parallel coordinate technology, the functions such as brush technology and axial sequencing were added to make a preliminary exploration of marine multi-discipline data, and the scatter matrix was introduced to show the distribution of each element. Secondly, the angle and area between data lines in parallel coordinates as well as the distance of element distribution in the scatter diagram were used as the difference measurement method to construct the similarity matrix for the calculated differences. Then the multidimensional scaling method was used to obtain the expression of the original data in the low dimensional space. Finally, K-means algorithm was used for clustering analysis of low-dimensional data after dimensionality reduction. The method proposed in this paper realizes data analysis and feature mining from a visual perspective, and obtains a is ualized display of high-dimensional data in a low-dimensional space, which effectively quantifies the correlation between different elements of marine data.
查看全文    下载PDF阅读器
关闭