CSForest, CSVoting and BCSForest

It appears you don't have a PDF plugin for this browser. No biggie... you can click here to download the PDF file.


This publication proposes a new cost-sensitive decision forest algorithm which we called CSForest. It's a combination of the core concepts behind CSTree and SysFor. CSForest is able to build an ensemble of cost-sensitive decision trees where each tree achieves high performance and uses the entire dataset. Thus, the models trained by CSForest are suitable for knowledge discovery and high performance machine learning. Also included are CSVoting: A method for classification using a decision forest, and BCSForest: an extension of CSForest that balances class distributions. This is my most highly cited research publication.