Title :
LogitBoost with Trees Applied to the WCCI 2006 Performance Prediction Challenge Datasets
Author :
Lutz, Roman Werner
Author_Institution :
ETH Zurich, Zurich
Abstract :
We apply LogitBoost with a tree-based learner to the five WCCI 2006 performance prediction challenge datasets. The number of iterations and the tree size is estimated by 10-fold cross-validation. We add a simple shrinkage strategy to make the algorithm more stable. The results are very promising since we won the challenge.
Keywords :
pattern classification; statistical analysis; trees (mathematics); LogitBoost; high-dimensional classification problems; prediction challenge datasets; shrinkage strategy; tree-based learner; Bit error rate; Error analysis; Logistics; Predictive models; Protection; Regression tree analysis; Seminars; Statistics; Terminology; Testing;
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.2006.246633
