聚类分析在主跃层识别中的应用

Applying of cluster analysis in main spring recognition

文章讨论了用有序样本聚类的方法来实现对主跃层的识别,文中建立了3种不同的分类函数和类直径的有序样本聚类模型。作者以中国海30分方区按月统计声速剖面历史数据为样本,分析了这3种不同的有序样本聚类方法的优劣。应用其中较优模型有效地提取了主跃层的厚度和平均梯度等信息,确定主跃层的起始点,有利于对声速剖面的特征提取,从而有效地对声速剖面进行分类。

The paper introduces a method of main spring recognition depending on a technique of sequential sample cluster and constructs three models with different classification functions and diameters. It also analyses the sound speed profile data which are statistic at monthly and 30'×30' latitude-longitude grid elements and compares the performance of three models. Using the best method the thickness, the start layer, and the average grad of main spring can be acquired. That is helpful to character abstract and classification of sound speed profile.