摘要
投影寻踪分类(PPC)模型的约束条件是∑p j=1a2j=1并且1≥aj≥-1。针对相同的密度窗宽R值,改变指标归一化方式前后PPC建模具有投影向量系数(权重)互为相反数和目标函数值保持不变等特性,据此提出建立正确的PPC模型的基本原则和步骤等,确保PPC建模得到真正的全局最优解和可靠结果,并进行实证建模研究。
The reasonable and correct constraint of projection vector coefficients (weights)is p∑j=1aj^2=1 and 1≥aj≥-1. For the same R value and the indexes with two different normalization modes, PPC model characters that the weights of in- dexes are opposite values and the objective function is constant. The proper principles and steps for establishing projection pursuit clustering model and the rules for judging indexes' attribute is put forward. The reasonable and reliable PPC model with real global optimal solutions is thus exploratory established and positive analyses are carried out.
出处
《科技管理研究》
CSSCI
北大核心
2014年第6期166-171,共6页
Science and Technology Management Research
基金
上海高校"工商管理"一流(培育)学科建设项目
上海市教委重点学科"商务经济学"建设项目
关键词
投影寻踪分类模型
投影向量系数(权重)
约束条件
全局最优解
projection pursuit clustering model
projection vector coefficients (weights)
constraint condition
global op-timal solutions