投影寻踪动态聚类模型

投影寻踪聚类模型在多因素聚类分析中被广泛应用并取得了满意的效果,然而,该模型还存在诸如密度窗宽参数取值经验确定等不足,有待改进提高.本文针对投影寻踪聚类模型的不足,首次把投影寻踪的思想和动态聚类方法结合起来构造投影指标,基于免疫进化算法,建立了投影寻踪动态聚类新模型.新模型一方面在整个运算过程中毋需人为给定参数,聚类结果客观、明确,另一方面,它还具有稳定性好、操作简便等特点.洪水分类的实际应用表明,投影寻踪动态聚类模型切实可行,取得了很好的效果,在多因素聚类分析领域具有广阔的应用前景.

Projection pursuit dynamic cluster model

Projection pursuit cluster(PPC) model is widely used in multifactor cluster analysis,but there are few problems needed to be solved in practice,and cutoff radius,which is an important parameter,is one of them.Aimed at the existing problems of PPC model and based on immune evolution algorithm,a projection pursuit dynamic cluster(PPDC) model,which combines dynamic cluster method with projection pursuit principle,is proposed for the first time in this study.As to PPDC model,the key of it lies in how to construct the projection index,and its operation process includes four steps.Compared with PPC model,PPDC model has its outstanding advantages,on one hand,it successfully avoids the problem of parameter calibration in PPC model,and on the other hand,the final cluster results can be output directly,and thus the subjectivity of PPC model is effectively eliminated.So PPDC model makes the cluster results more objective and definite,and besides that,it is also robust and easy to operate in practice.PPDC model is applied to study flood classification,and the test shows that PPDC model is feasible and effective,and it provides a powerful tool for the solution of similar problems.