mabelc / ssc

The ssc R package

ssc

Provides a collection of self-labeled techniques for semi-supervised classification. In semi-supervised classification, both labeled and unlabeled data are used to train a classifier. This learning paradigm has obtained promising results, specifically in the presence of a reduced set of labeled examples. This package implements a collection of self-labeled techniques to construct a classification model. These techniques enlarges the original labeled set using the most confident predictions to classify unlabeled data. The techniques implemented can be applied to classification problems in several domains by the specification of a suitable supervised base classifier. At low ratios of labeled data, it can be shown to perform better than classical supervised classifiers.

Installation

You can install ssc from github with:

# install.packages("devtools")
devtools::install_github("mabelc/SSC")

Example

This is a basic example which shows you how to solve a common problem:

library(ssc)

## Load Iris data set
data(iris)

x <- iris[, -5] # instances without classes
x <- as.matrix(x)
y <- iris$Species

## Prepare data
set.seed(1)
# 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: Training from a set of instances with 1-NN as base classifier.
library(caret)
#> Loading required package: lattice
#> Loading required package: ggplot2
m <- selfTraining(x = xtrain, y = ytrain, learner = knn3, learner.pars = list(k = 1))
pred <- predict(m, xitest)
table(pred, yitest)
#>             yitest
#> pred         setosa versicolor virginica
#>   setosa         24          0         0
#>   versicolor      0         22         9
#>   virginica       0          0        20

Overview of the package

The ssc package implements the following self-labeled techniques:

• coBC
• democratic
• selfTraining
• setred
• snnrce
• triTraining

selfTraining, setred and snnrce are single-classifier based. coBC, democratic and triTraining are multi-classifier based.

ssc's Issues

About coBC and coBCG

Thank you for the package, I find it very useful.

As for the latest version, I'd like to clarify about coBC and coBCG functions, do these implement co-bagging by Hady M, Schwenker F (2008)? Because in the R documentation, the reference is Blum A, Mitchell T (1998), while in the detailed documentation here https://cran.r-project.org/web/packages/ssc/vignettes/ssc.pdf, it seems to suggest that co-bagging by Hady M, Schwenker F (2008) is implemented in for the functions. So, I would like to clarify whether co-bagging was implemented instead of the basic Blum A, Mitchell T (1998).

Thank you.

Method to retrieve predictions from SSC

Dear developers,

I was looking for a Semi-Supervised ML method in R and found your excellent package. I tried your example code adapting it to my input data, and after some reformating it works apparently well. The problem I have is related to how to access prediction results for each of the rows in my input table.
I may sound naive, but I can't find the code to access the classification assigned for each of the "unlabeled" rows in my table, by any of the methods carried out in your vignette's example code.
I can access the sumary of how many samples have been assigned to each class, but I'd like to know how to access to each row's individual class/label prediction (in dataframe format, for instance).
I hope I was able to explain myself clearly enough for everybody to understand this request.
Thanks in advance and congrats for your nice work.

Return probability with selfTraining and SVM

I'm trying to return the class probabilities as predicted by selfTraining and SVM. The code below is from the selfTraining documentation. It seems like it should be returning probability but is not. Any help?

## Load Wine data set
data(wine)
cls <- which(colnames(wine) == "Wine")
x <- wine[, -cls] # instances without classes
y <- wine[, cls] # 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: Training from a set of instances with 1-NN as base classifier.
learner <- e1071::svm
learner.pars <- list(type = "C-classification", kernel="radial", 
                     probability = TRUE, scale = TRUE)
pred <- function(m, x){
  r <- predict(m, x, probability = TRUE)
  prob <- attr(r, "probabilities")
  prob
}
m3 <- selfTraining(x = xtrain, y = ytrain, 
                   learner = learner, 
                   learner.pars = learner.pars, 
                   pred = pred)
pred3 <- predict(m3, xitest, probability=TRUE)
head(attr(pred3, "probabilities"))