投影寻踪聚类模型及其应用  

Projection pursuit cluster model and its application

作　　者：王顺久[1] 张欣莉[2] 丁晶[1] 侯玉[3] 

机构地区：[1]四川大学水电学院,四川成都610065 [2]四川大学工商管理学院,四川成都610064 [3]中山大学地理学系,广东广州510275

出　　处：《长江科学院院报》2002年第6期53-55,61,共4页Journal of Changjiang River Scientific Research Institute

基　　金：国家自然科学基金资助项目(50079031)

摘　　要：投影寻踪聚类模型已广泛应用于涉及多因素的综合评价问题,但模型中密度窗宽参数的确定目前还没有一种很好的方法,现主要通过试算或经验来确定,缺乏理论根据。针对这一问题,在较为详细介绍模型结构的基础上,从模型的数学构造入手对密度窗宽参数的取值进行了分析探讨,推导得出了计算的经验公式,实际应用表明该方法切实可行,为模型的推广应用奠定了基础,同时也为涉及多因素的方案决策提供了一条新途径。Projection pursuit cluster model has been widelly used in assessing water quality, selecting analogy basins, evaluating flood disasters, and other similar problems. Projection pursuit can find the main characteristics in low dimension space, and then a right result can be got according to the main characteristic.But the density window parameter R in the model is determined by trial and error method or experience lacking theoretical basis. Aiming at this point, based on describing in detailed model structure, initiating from models mathematical construction an empirical formula for calculating R is derived.The application results show that this method is reasonable, feasible in practice. It is a new way for solving multi?factor decision problems.

关 键 词：投影寻踪 聚类分析 河流 分类 

分 类 号：TV212[水利工程—水文学及水资源]