ssc (version 1.0)

setred: Train the SETRED model

Description

Builds and trains a model to predict the label of instances, according to SETRED algorithm.

Usage

setred(x, y, bclassif = bClassifOneNN(), dist = "matrix", theta = 0.1, min.amount = ceiling(length(which(is.na(y))) * 0.3), max.iter = 50)

Arguments

x

A object that can be coerced as matrix. This object have various interpretations depending on the value set in dist argument. See dist argument.

y

A vector with the labels of training instances. In this vector the unlabeled instances are specified with the value NA.

bclassif

Base classifier specification. Default is bClassifOneNN(). For defining new base clasifiers see bClassif.

dist

Distance information. Valid options are:

• "matrix": this string indicates that x is a distance matrix.
• string: the name of a distance method available in proxy package. In this case x is interpreted as a matrix of instances.
• function: a function defined by the user that computes the distance between two vectors. This function is called passing the vectors in the firsts two arguments. If the function have others arguments, those arguments must be have default values. In this case x is interpreted as a matrix of instances.

theta

Rejection threshold to test the critical region. Default is 0.1.

min.amount

Minimum number of unlabeled instances for stopping training. When the amount of remaining instances reaches this number the self-labeling process is stopped. Default is 0.3 * .

max.iter

Maximum number of iterations in the self-labeling process. Default is 50.

Value

The trained model stored in a list with the following named values:

References

Ming Li and ZhiHua Zhou. Setred: Self-training with editing. In Advances in Knowledge Discovery and Data Mining, volume 3518 of Lecture Notes in Computer Science, pages 611–621. Springer Berlin Heidelberg, 2005. ISBN 978-3-540-26076-9. doi: 10.1007/11430919 71.

Examples

# This example is part of SETRED demo.
# Use demo(SETRED) to see all the examples.

## Load Wine data set
data(wine)

x <- wine[, -14] # instances without classes
y <- wine[, 14] # the classes
x <- scale(x) # scale the attributes

## Prepare data
set.seed(20)
# Use 50% of instances for training
tra.idx <- sample(x = length(y), size = ceiling(length(y) * 0.5))
xtrain <- x[tra.idx,] # training instances
ytrain <- y[tra.idx]  # classes of training instances
# Use 70% of train instances as unlabeled set
tra.na.idx <- sample(x = length(tra.idx), size = ceiling(length(tra.idx) * 0.7))
ytrain[tra.na.idx] <- NA # remove class information of unlabeled instances

# Use the other 50% of instances for inductive testing
tst.idx <- setdiff(1:length(y), tra.idx)
xitest <- x[tst.idx,] # testing instances
yitest <- y[tst.idx] # classes of testing instances

## Example: Using the Euclidean distance in proxy package.
m <- setred(xtrain, ytrain, dist = "Euclidean")
pred <- predict(m, xitest)
caret::confusionMatrix(table(pred, yitest))