沉积物污染特征分析的未确知聚类-对应分析模型

Coupled Model of Unascertained Cluster and Correspondence Factor Analysis for Pollution Characteristics of Sediments

针对现有未确知聚类的不足,将聚类有效性函数引入聚类算法中,通过与对应分析方法的综合集成,建立了未确知聚类-对应分析(UC-CFA)模型,并以加速遗传算法优化计算最佳聚类数目、聚类中心以及聚类点(包括污染指标和采样点)隶属于聚类中心的隶属度.以安徽巢湖塘西河口湿地为研究对象,采用UC-CFA模型对沉积物污染特征进行聚类分析,将所有聚类点划分为4类,即Cd-有机质联合污染区,高磷污染区,Cu-Zn-Pb-Cr复合污染区以及高氮污染区等.实例研究表明,对于沉积物污染特征分析,UC-CFA模型具有适用性.

Aiming at the deficiencies of present unascertained cluster theory, a clustering validity function was introduced to the unascertained C-means cluster algorithm. On the basis of synthesizing unascertained cluster （UC） and correspondence factor analysis （CFA）, a coupled model, i.e. UC-CFA, was established for pollution characteristic analysis. According to the composite model, such elements as optimal cluster number, cluster centers and membership degrees of clustering points （including all pollution indices and sampling points） subordinating to corresponding cluster centers can be optimized and computed by means of accelerating genetic algorithm. As a case, the UC-CFA model was employed for the pollution analysis of sediments in the estuarine wetland of Tangxihe River, a main inflow stream of Lake Caohu. As a result, all the clustering points were roughly divided into four groups, namely united contaminated area of Cd-organic matter, high phosphorus contaminated area, united contaminated area of Cu-Zn-Pb-Cr and high-nitrogen contaminated area, etc. The results show that it is suitable for the UC-CFA model to conduct the pollution characteristic analysis of river sediments.