投影寻踪建模中关键参数合理值的确定与分析

计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (9): 50-55.

投影寻踪建模中关键参数合理值的确定与分析

熊聘1，楼文高1，2

1.上海理工大学光电信息与计算机工程学院，上海 200093
2.上海商学院，上海 200235

出版日期:2016-05-01 发布日期:2016-05-16

Determination and analysis of reasonable value of key parameter in projection pursuit clustering modelling

XIONG Pin1, LOU Wengao1，2

1.School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China
2.Shanghai Business School, Shanghai 200235, China

Online:2016-05-01 Published:2016-05-16

摘要/Abstract

摘要： 从理论和实证两个视角分析了四种选取合理的窗宽半径[R]值方案对投影寻踪分类（PPC）建模结果的影响，指出了投影寻踪分类建模中存在的问题，研究结果表明：[R]取较小值[(R0.1Sy)]方案时，通常情况下最优化过程无法求得真正的全局最优解;[R]取较大值[(rmaxR2m)]方案，不仅其证明过程是错误的，而且与PPC建模的基本思想相矛盾;[R]取常数的方案，也与PPC建模选取[R]值的要求不一致;只有[R]取中间适度值[(rmax/5Rrmax/3)]方案是合理和正确的，既符合PPC建模的基本思想和选取[R]值的要求，最优化过程也能求得真正的全局最优解。

关键词: 投影寻踪分类建模, 窗宽半径[R]值, 合理的[R]值范围, 中间适度值方案

Abstract: Projection Pursuit Clustering（PPC） technique has been widely used in various fields.The influence of four strategies for determining the range of Cutoff Radius R Value（CRRV） on the PPC modelling is theoretically and practically discussed and analyzed in this paper.The result indicates that the smaller CRRV[(R0.1Sy)] cannot reach the global optimum in general; in terms of the larger CRRV[(rmaxR2m)], not only its proof is wrong, but cannot meet the concept of PPC modeling, although the global optimal solution can be obtained; the constant of CRRV is unreasonable and incorrect because of contradiction on requirement of [R] value; the moderate-suitable CRRV[(rmax/5Rrmax/3)], however, is the reasonable and correct strategy because of the match of PPC modelling concept and requirement for [R] value.

Key words: Projection Pursuit Clustering（PPC） modelling, cutoff radius [R] value, reasonable [R] value, strategy of taking moderate-suitable [R] value

熊聘1，楼文高1，2. 投影寻踪建模中关键参数合理值的确定与分析[J]. 计算机工程与应用, 2016, 52(9): 50-55.

XIONG Pin1, LOU Wengao1，2. Determination and analysis of reasonable value of key parameter in projection pursuit clustering modelling[J]. Computer Engineering and Applications, 2016, 52(9): 50-55.