plotcoords: Plot smvcir discriminant coordinates

In danno11/SMVCIR: Sliced Mean Variance Covariance Inverse Regression

Description

Plots specified dimensions of smvcir space.

Usage

plotcoords(x, dimensions = c(1, 2), groups = 1:10, GL = FALSE,
  ID = FALSE, build_svm = FALSE, kernel = NULL, svmModel = NULL)

Arguments

x	Either an smvcir model object or a data frame of discriminant coordinates.
dimensions	Dimensions to be plotted. Must have three if a 3d plot is desired. Otherwise pairwise plots are produced.
GL	If GL = TRUE and three dimensions above are specified, a three-dimensional plot using the rgl package will be produced.
build_svm	If build_svm=TRUE, a support vector machine model is built using the dimensions specificied as covariates, and can be visualized with respect to the plotted dimensions.
svmModel	If desired, this option may be used to plot new coordinates along with an svm model built from other data coordinates (possibly desirable if training/testing splits are used). The svmModel specified must also use the same covariates as those to be plotted. See example below.

Examples

library(caret)
train<-createDataPartition(pima$diabetes, p = .8, list = FALSE)
pim.smv<-smvcir("diabetes", data = pima[train,], test = T) ###Build smvcir model on training set
pred_coords<-predict(pim.smv, newdata = pima[-train,], coordinates_only = TRUE)
pred_coords$diabetes<-pima$diabetes[-train]

plotcoords(pred_coords, dimensions = 1:8)
plotcoords(pred_coords, dimensions = 1:3, GL = TRUE, build_svm = TRUE)
plotcoords(pred_coords, dimensions = 1:3, build_svm = TRUE)

data(banknote)
###Create training rows for Banknote data
train<-sample(1:nrow(banknote), nrow(banknote)*.8, replace = FALSE)
####Build smvcir model on training set
banksmv<-smvcir("Y", data = banknote[train,], test = TRUE)
####Build svm model on first three dimensions of smvcir discriminant coordinates
svmModel<-svm(Y~., data = banksmv$direct[,c(1:3,7) ], probability = TRUE)

###Get discriminant coordinates from test set
bcoords<-predict(banksmv, newdata = banknote[-train,], coordinates_only = TRUE, maxdim = 3)
bcoords$Y<-banknote[-train,]$Y

####Plot test set coordinates with respect to svm model built from training set
plotcoords(bcoords, dimensions = 1:3, GL = TRUE, build_svm = TRUE, svmModel = svmModel)

danno11/SMVCIR documentation built on May 14, 2019, 6:06 p.m.