FLXMCLsvm: Mixtures of Support Vector Machines

In locClass: Collection of Local Classification Methods

Description

This is a model driver for flexmix implementing mixtures of Support Vector Machines.

Usage

  FLXMCLsvm(formula = . ~ ., ...)

Arguments

formula	A formula which is interpreted relative to the formula specified in the call to flexmix using update.formula. Only the left-hand side (response) of the formula is used. Default is to use the original flexmix model formula.
...	Further arguments to and from other methods.

Value

Returns an object of class FLXMCLsvm inheriting from FLXMCL.

Examples

library(locClassData)
data <- flashData(1000)
grid <- expand.grid(x.1=seq(-6,6,0.2), x.2=seq(-4,4,0.2))

cluster <- kmeans(data$x, center = 2)$cluster
model <- FLXMCLsvm(kernel = "linear")
fit <- flexmix(y ~ ., data = as.data.frame(data), concomitant = FLXPwlda(~ x.1 + x.2), model = model, cluster = cluster, control = list(classify = "hard"))

## prediction for single component models without aggregation
pred.grid <- predict(fit, newdata = grid)
image(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]["decision",][[1]][,1], length(seq(-6,6,0.2))))
contour(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]["decision",][[1]][,1], length(seq(-6,6,0.2))), add = TRUE)
points(data$x, pch = as.character(data$y))

image(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[2]]["decision",][[1]][,1], length(seq(-6,6,0.2))))
contour(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[2]]["decision",][[1]][,1], length(seq(-6,6,0.2))), add = TRUE)
points(data$x, pch = as.character(data$y))

## prediction with aggregation depending on membership in mixture components
pred.grid <- mypredict(fit, newdata = grid, aggregate = TRUE)
image(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]$decision[,1], length(seq(-6,6,0.2))))
contour(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]$decision[,1], length(seq(-6,6,0.2))), add  = TRUE)
points(data$x, pch = as.character(data$y))

image(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]$posterior[,1], length(seq(-6,6,0.2))))
contour(seq(-6,6,0.2), seq(-4,4,0.2), matrix(pred.grid[[1]]$posterior[,1], length(seq(-6,6,0.2))), add  = TRUE)
points(data$x, pch = as.character(data$y))

## local memberhsip
loc.grid <- prior(fit, newdata = grid)
contour(seq(-6,6,0.2), seq(-4,4,0.2), matrix(loc.grid[,1], length(seq(-6,6,0.2))), add  = TRUE)

locClass documentation built on May 2, 2019, 5:21 p.m.