Could someone give a better example of how to extend [1] to create an
ensemble from multiple machine learning classifiers to improve the
classification performance?
[1]
http://heuristically.wordpress.com/2009/12/23/compare-performance-machine-learning-classifiers-r/
Berry and Linoff ("Mastering Data Mining", p217) suggest the approach
below, but the performance is similar to the individual models.
# create ensemble (per Berry and Linoff)
evidence_a <- attr(x.svm.prob, "probabilities")[,2] * x.ip.prob[,2] *
x.cf.prob
evidence_b <- ( 1-x.ip.prob[,2]) * (1-attr(x.svm.prob,
"probabilities")[,2] * (1-x.cf.prob))
x.en.prob <- (evidence_a) / (evidence_a + evidence_b)
# bagging
x.ip.prob.rocr <- prediction(x.ip.prob[,2], BreastCancer[ind == 2,'Class'])
x.ip.perf <- performance(x.ip.prob.rocr, "tpr","fpr")
plot(x.ip.perf, col=2)
# svm
x.svm.prob.rocr <- prediction(attr(x.svm.prob, "probabilities")[,2],
BreastCancer[ind == 2,'Class'])
x.svm.perf <- performance(x.svm.prob.rocr, "tpr","fpr")
plot(x.svm.perf, col=3, add=TRUE)
# cforest
x.cf.prob.rocr <- prediction(x.cf.prob, BreastCancer[ind == 2,'Class'])
x.cf.perf <- performance(x.cf.prob.rocr, "tpr","fpr")
plot(x.cf.perf, col=4, add=TRUE)
# ensemble
#x.en.rocr <- prediction(rowMeans(cbind(x1=x.cf.prob,
x2=x.ip.prob[,2])), BreastCancer[ind == 2,'Class'])
x.en.rocr <- prediction(x.en.prob, BreastCancer[ind == 2,'Class'])
x.en.perf <- performance(x.en.rocr, "tpr","fpr")
plot(x.en.perf, col=5, add=TRUE)
# Draw a legend.
legend(0.6, 0.6, c('bagging','svm', 'cforest', 'ensemble'), 2:5)
Jack
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
