Title :
Classification with Tree-Based Ensembles Applied to the WCCI 2006 Performance Prediction Challenge Datasets
Author :
Dahinden, Corinne
Author_Institution :
ETH Zurich, Zurich
Abstract :
Our contribution to the WCCI 2006 performance prediction challenge is built on a modified random forests scheme, with cross-validation as a means for tuning parameters and estimating error-rates. This simple and computationally very efficient approach was found to yield better predictive performance than many algorithms of much higher complexity.
Keywords :
error statistics; pattern classification; trees (mathematics); WCCI 2006 Performance Prediction Challenge datasets; error-rate estimation; modified random forests scheme; tree-based ensembles; Classification algorithms; Classification tree analysis; Error analysis; High performance computing; Humans; Input variables; Machine learning algorithms; Parameter estimation; Regression tree analysis; Seminars;
Conference_Titel :
Neural Networks, 2006. IJCNN '06. International Joint Conference on
Conference_Location :
Vancouver, BC
Print_ISBN :
0-7803-9490-9
DOI :
10.1109/IJCNN.2006.246635
