PPtree (version 2.3.0)

PP.optimize: Find optimal projection by maximizing selected PP index

Description

Find optimal projection using PP index.

Usage

PP.optimize.random(PPmethod, projdim, data, class, std=TRUE, cooling=0.99, temp=1, r=NULL, lambda=NULL, weight=TRUE, ...)
PP.optimize.anneal(PPmethod, projdim, data, class, std=TRUE, cooling=0.999, temp=1, energy=0.01, r=NULL, lambda=NULL, weight=TRUE, ...)
PP.optimize.Huber(PPmethod, projdim, data, class, std=TRUE, cooling=0.99, temp=1, r=NULL, lambda=NULL, weight=TRUE, ...)
PP.optimize.plot(PP.opt, data, class, std=TRUE)

Arguments

PPmethod

Selected PP index

``LDA" - LDA index

``Lp" - Lp index;

``PDA" - PDA index

projdim

dimension of projection that you want to find

data

data without class information

class

class information

std

decide whether data will be standardized or not before applying projection pursuit

weight

weight flag using in LDA index

cooling

parameter for optimization

temp

inital temperature for optimization

energy

parameter for simulated annealing optimization

r

a parameter for $L_r$ index

lambda

a parameter for PDA index

PP.opt

the optimal projection

...

...

Value

index.best

PP index for optimal projected data

proj.best

optimal projection

References

Lee E., Cook D., and Klinke, S. (2002) Projection Pursuit indices for supervised classification

See Also

{PPindex.class}

Examples

data(iris)

PP.opt<-PP.optimize.random("LDA",1,iris[,1:4],iris[,5],cooling=0.999,temp=1)

PP.opt$index.best
PP.optimize.plot(PP.opt,iris[,1:4],iris[,5])


PP.opt<-PP.optimize.anneal("LDA",1,iris[,1:4],iris[,5],cooling=0.999,temp=1,energy=0.01)
PP.opt$index.best

PP.optimize.plot(PP.opt,iris[,1:4],iris[,5])


PP.opt<-PP.optimize.Huber("LDA",2,iris[,1:4],iris[,5],cooling=0.999,r=1)
PP.opt$index.best
PP.optimize.plot(PP.opt,iris[,1:4],iris[,5])