| Title: | Projection Pursuit Classification Tree Visualization |
| Description: | Tools for exploring projection pursuit classification tree using various projection pursuit indexes. |
| Version: | 2.0.4 |
| Author: | Eun-Kyung Lee |
| Maintainer: | Eun-Kyung Lee <lee.eunk@gmail.com> |
| Date: | 2019-12-6 |
| Depends: | R (≥ 2.10), gridExtra, ggplot2, partykit, Rcpp |
| Suggests: | MASS (≥ 3.1-20), stats |
| LinkingTo: | Rcpp, RcppArmadillo |
| License: | GPL-2 | GPL-3 [expanded from: GPL (≥ 2)] |
| RoxygenNote: | 7.0.2 |
| NeedsCompilation: | yes |
| Packaged: | 2019-12-06 07:44:42 UTC; EKLee |
| Repository: | CRAN |
| Date/Publication: | 2019-12-06 08:40:02 UTC |
ENTROPY PPindex
Description
ENTROPY projection pursuit index
Usage
ENTROPYindex1D(origclass,origdata,proj)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
proj |
projection matrix |
Details
Calculate ENTROPY projection pursuit index value
Examples
data(iris)
ENTROPYindex1D(iris[,5],as.matrix(iris[,1,drop=FALSE]))
GINI PPindex
Description
GINI projection pursuit index
Usage
GINIindex1D(origclass,origdata,proj)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
proj |
projection matrix |
Details
Calculate GINI projection pursuit index value
Examples
data(iris)
GINIindex1D(iris[,5],as.matrix(iris[,1,drop=FALSE]))
Huber plot
Description
Huber plot for 2D data
Usage
Huberplot(origdata2D,origclass,PPmethod="LDA",weight=TRUE,r=1,
lambda=0.5,opt.proj=TRUE,UserDefFtn=NULL,...)
Arguments
origdata2D |
2-dimensional numerical data for Huber plot |
origclass |
class information vector of data |
PPmethod |
method for projection pursuit; "LDA", "PDA", "Lr", "GINI", "ENTROPY", and "UserDef" |
weight |
weight flag in LDA, PDA and Lr index |
r |
r in Lr index |
lambda |
lambda in PDA index |
opt.proj |
flag to show the best projection in the plot |
UserDefFtn |
User defined index function when PPmethod="UserDef" |
... |
arguments to be passed to methods |
Details
Draw Huber plot for 2-dimensional data with various PP indices and the histogram of the projected data onto the optimal projection to explore the behavior of the projection prsuit indices
References
Lee, EK., Cook, D., Klinke, S., and Lumley, T.(2005) Projection Pursuit for Exploratory Supervised Classification, Journal of Computational and Graphical Statistics, 14(4):831-846.
Examples
data(iris)
Huberplot(iris[,1:2],iris[,5],PPmethod="LDA")
LDA PPindex
Description
LDA projection pursuit index
Usage
LDAindex(origclass,origdata,proj,weight=TRUE)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
proj |
projection matrix |
weight |
weight flag in LDA |
Details
Calculate LDA projection pursuit index value
References
Lee, EK., Cook, D., Klinke, S., and Lumley, T.(2005) Projection Pursuit for exploratory supervised classification, Journal of Computational and Graphical statistics, 14(4):831-846.
Examples
data(iris)
LDAindex(iris[,5],as.matrix(iris[,1:4]))
PP optimization using LDA index
Description
PP optimization using LDA index
Usage
LDAopt(origclass,origdata,q=1,weight=TRUE,...)
Arguments
origclass |
class information vector of data |
origdata |
data matrix without class information |
q |
dimension of projection vector |
weight |
weight flag in LDA index |
... |
arguments to be passed to methods |
Details
Find the q-dimensional optimal projection using LDA projectin pursuit index
Value
indexbest maximum LDA index value
projbest optimal q-dimensional projection matrix
origclass original class information vector
origdata original data matrix without class information
References
Lee, EK., Cook, D., Klinke, S., and Lumley, T.(2005) Projection Pursuit for Exploratory Supervised Classification, Journal of Computational and Graphical Statistics, 14(4):831-846.
Examples
data(iris)
LDA.proj.result <- LDAopt(iris[,5],iris[,1:4])
LDA.proj.result$indexbest
LDA.proj.result$projbest
Lr PPindex
Description
Lr projection pursuit index
Usage
Lrindex(origclass,origdata,proj,weight=TRUE,r=1)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
proj |
projection matrix |
weight |
weight flag in Lr index |
r |
r in Lr index |
Details
Calculate Lr projection pursuit index value
References
Lee, EK., Cook, D., Klinke, S., and Lumley, T.(2005) Projection Pursuit for exploratory supervised classification, Journal of Computational and Graphical statistics, 14(4):831-846.
Examples
data(iris)
Lrindex(iris[,5],as.matrix(iris[,1:4]),r=1)
PDA PPindex
Description
PDA projection pursuit index
Usage
PDAindex(origclass,origdata,proj,weight=TRUE,lambda=0.1)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
proj |
projection matrix |
weight |
weight flag in PDA |
lambda |
lambda in PDA index |
Details
Calculate PDA projection pursuit index value
References
Lee, EK., Cook, D.(2010) A projection pursuit index for large p small n data, Statistics and Computing, 20:381-392.
Examples
data(iris)
PDAindex(iris[,5],as.matrix(iris[,1:4]),lambda=0.2)
PP optimization using PDA index
Description
PP optimization using PDA index
Usage
PDAopt(origclass,origdata,q=1,weight=TRUE,lambda=0.1,...)
Arguments
origclass |
class information vector of data |
origdata |
data matrix without class information |
q |
dimension of projection vector |
weight |
weight flag in PDA index |
lambda |
lambda in PDA index |
... |
arguments to be passed to methods |
Details
Find the q-dimensional optimal projection using PDA projectin pursuit index
Value
indexbest maximum PDA index value
projbest optimal q-dimensional projection matrix
origclass original class information vector
origdata original data matrix without class information
References
Lee, EK, Cook, D.(2010) A Projection Pursuit Index for Large p Small n Data, Statistics and Computing, 20:381-392.
Examples
data(iris)
PDA.proj.result <- PDAopt(iris[,5],iris[,1:4],weight=TRUE,q=2,lambda=0.1)
PDA.proj.result$indexbest
PDA.proj.result$projbest
Projection pursuit classification tree
Description
Construct the projection pursuit classification tree
Usage
PPTreeclass(formula,data, PPmethod="LDA",weight=TRUE,
r=1,lambda=0.1,energy=0,maxiter=50000,...)
Arguments
formula |
an object of class "formula" |
data |
data frame |
PPmethod |
method for projection pursuit; "LDA", "PDA", "Lr", "GINI", and "ENTROPY" |
weight |
weight flag in LDA, PDA and Lr index |
r |
r in Lr index |
lambda |
lambda in PDA index |
energy |
parameter for the probability to take new projection |
maxiter |
maximum iteration number |
... |
arguments to be passed to methods |
Details
Find tree structure using various projection pursuit indices of classification in each split.
Value
Tree.Struct tree structure of projection pursuit classification tree
projbest.node 1 dimensional optimal projections of each node split
splitCutoff.node cutoff values of each node split
origclass original class
origdata original data
References
Lee, YD, Cook, D., Park JW, and Lee, EK(2013) PPtree: Projection Pursuit Classification Tree, Electronic Journal of Statistics, 7:1369-1386.
Examples
data(iris)
Tree.result <- PPTreeclass(Species~.,data = iris,"LDA")
Tree.result
PPtree node visualization
Description
Visualization tools to explore each node of PPtree
Usage
PPclassNodeViz(PPclassOBJ,node.id,Rule,legend,std,image,diff.prop)
Arguments
PPclassOBJ |
PPregclass object |
node.id |
node ID |
Rule |
cutoff rule |
legend |
flag to represent legend in the plot. Default value is TRUE |
std |
flag to standardize data before drawing plot |
image |
flag to draw image plot of correlation matrix |
diff.prop |
percentage of number of variables with significant differences and shown in red in the bar chart style means |
Details
For the inner node, four plots are provided - the bar chart style plot with projection pursuit coefficients of each variable, the histogram of the projected data, the bar chart style plots with means of each variables for the left and the right group, and the image plot of correlations.
References
Lee, YD, Cook, D., Park JW, and Lee, EK(2013) PPtree: Projection Pursuit Classification Tree, Electronic Journal of Statistics, 7:1369-1386.
Examples
data(iris)
Tree.result <- PPTreeclass(Species~., data = iris,"LDA")
Tree.result
PPclassNodeViz(Tree.result,1,1)
predict PPtree
Description
predict projection pursuit classification tree
Usage
PPclassify(Tree.result,test.data,Rule,true.class=NULL,...)
Arguments
Tree.result |
PPtreeclass object |
test.data |
the test dataset |
Rule |
split rule 1: mean of two group means 2: weighted mean of two group means - weight with group size 3: weighted mean of two group means - weight with group sd 4: weighted mean of two group means - weight with group se 5: mean of two group medians 6: weighted mean of two group medians - weight with group size 7: weighted mean of two group median - weight with group IQR 8: weighted mean of two group median - weight with group IQR and size |
true.class |
true class of test dataset if available |
... |
arguments to be passed to methods |
Details
Predict class for the test set with the fitted projection pursuit classification tree and calculate prediction error.
Value
predict.class predicted class
predict.error number of the prediction errors
References
Lee, YD, Cook, D., Park JW, and Lee, EK(2013) PPtree: Projection Pursuit Classification Tree, Electronic Journal of Statistics, 7:1369-1386.
Examples
data(iris)
n <- nrow(iris)
tot <- c(1:n)
n.train <- round(n*0.9)
train <- sample(tot,n.train)
test <- tot[-train]
Tree.result <- PPTreeclass(Species~.,data=iris[train,],"LDA")
PPclassify(Tree.result,iris[test,1:4],1,iris[test,5])
Projection pursuit optimization
Description
PP optimization using various projection pursuit indices
Usage
PPopt(origclass,origdata,q=1,PPmethod="LDA",weight=TRUE,r=1,
lambda=0.1,energy=0,cooling=0.999,TOL=0.0001,maxiter = 50000)
Arguments
origclass |
class information vector |
origdata |
data matrix without class information |
q |
dimension of projection matrix |
PPmethod |
method for projection pursuit; "LDA", "PDA", "Lr", "GINI", and "ENTROPY" |
weight |
weight flag in LDA, PDA and Lr index |
r |
r in Lr index |
lambda |
lambda in PDA index |
energy |
energy parameter |
cooling |
cooling parameter |
TOL |
tolerance |
maxiter |
number of maximum iteration |
Details
Find the q-dim optimal projection using various projectin pursuit indices with class information
Value
indexbest maximum LDA index value
projbest optimal q-dim projection matrix
origclass original class information vector
origdata original data matrix without class information
References
Lee, EK., Cook, D., Klinke, S., and Lumley, T.(2005) Projection Pursuit for exploratory supervised classification, Journal of Computational and Graphical statistics, 14(4):831-846.
Examples
data(iris)
PP.proj.result <- PPopt(iris[,5],as.matrix(iris[,1:4]))
PP.proj.result
PPopt visualization
Description
Visualize PPopt result
Usage
PPoptViz(PPoptOBJ)
Arguments
PPoptOBJ |
PPoptim object. result from LDAopt, PDAopt, and PPopt |
Details
Visualize the result of projection pursuit optimization
References
Lee, EK(2017) PPtreeViz: An R Package for Visualizing Projection Pursuit Classification Trees, Journal of Statistical Software <doi:10.18637/jss.v083.i08>
Examples
data(iris)
PPoptViz(LDAopt(iris[,5],iris[,1:4],q=1))
PPoptViz(LDAopt(iris[,5],iris[,1:4],q=2))
PPtree plot
Description
projection pursuit classification tree plot
Usage
## S3 method for class 'PPtreeclass'
plot(
x,
font.size = 17,
width.size = 1,
main = "Projection Pursuit Classification Tree",
sub = NULL,
...
)
Arguments
x |
PPtreeclass object |
font.size |
font size of plot |
width.size |
size of eclipse in each node. |
main |
main title |
sub |
sub title |
... |
arguments to be passed to methods |
Details
Draw projection pursuit classification tree with tree structure. It is modified from a function in party library.
References
Lee, EK(2017) PPtreeViz: An R Package for Visualizing Projection Pursuit Classification Trees, Journal of Statistical Software <doi:10.18637/jss.v083.i08>
Examples
data(iris)
Tree.result <- PPTreeclass(Species~., data = iris,"LDA")
Tree.result
plot(Tree.result)
predict PPtree
Description
predict projection pursuit classification tree
Usage
## S3 method for class 'PPtreeclass'
predict(object, newdata = NULL, Rule = 1, ...)
Arguments
object |
a fitted object of class inheriting from "PP.Tree.class" |
newdata |
the test dataset |
Rule |
split rule 1: mean of two group means 2: weighted mean of two group means - weight with group size 3: weighted mean of two group means - weight with group sd 4: weighted mean of two group means - weight with group se 5: mean of two group medians 6: weighted mean of two group medians - weight with group size 7: weighted mean of two group median - weight with group IQR 8: weighted mean of two group median - weight with group IQR and size |
... |
arguments to be passed to methods |
Details
Predict class for the test set with the fitted projection pursuit classification tree and calculate prediction error.
References
Lee, YD, Cook, D., Park JW, and Lee, EK(2013) PPtree: Projection Pursuit Classification Tree, Electronic Journal of Statistics, 7:1369-1386.
Examples
data(iris)
n <- nrow(iris)
tot <- c(1:n)
n.train <- round(n*0.9)
train <- sample(tot,n.train)
test <- tot[-train]
Tree.result <- PPTreeclass(Species~., data=iris[train,],"LDA")
predict(Tree.result)
Print PP.Tree.class result
Description
Print PP.Tree.class result
Usage
## S3 method for class 'PPtreeclass'
print(x, coef.print = FALSE, cutoff.print = FALSE, verbose = TRUE, ...)
Arguments
x |
PPtreeclass object |
coef.print |
print projection coefficients in each node ifTRUE |
cutoff.print |
print cutoff values in each node if TRUE |
verbose |
print if TRUE, no output if FALSE |
... |
arguments to be passed to methods |
Details
Print the projection pursuit classification tree result
References
Lee, EK(2017) PPtreeViz: An R Package for Visualizing Projection Pursuit Classification Trees, Journal of Statistical Software <doi:10.18637/jss.v083.i08>
Examples
data(iris)
Tree.result <- PPTreeclass(Species~.,data = iris,"LDA")
Tree.result
print(Tree.result,coef.print=TRUE,cutoff.print=TRUE)