Title: Projection pursuit supervised classification
Authors: Natalia da Silva - Universidad de la Republica (Uruguay) [presenting]
Di Cook - Monash University (Australia)
Eun-Kyung Lee - Ewha Womans University (Korea, South)
Abstract: Projection pursuit random forest (PPF) is a new ensemble learning method for classification problems, built from trees utilizing combinations of predictors. PPF builds a forest from many projection pursuit trees (PPtree); trees are constructed by splitting on linear combinations of randomly chosen variables. Projection pursuit is used to find the linear combination of variables that best separates groups, and many different rules to make the actual split are provided. Utilizing linear combinations of variables to separate classes takes the correlation between variables into account, which allows PPF to outperform a traditional random forest when separations between groups occur in combinations of variables. PPF can be used in multi-class problems and is implemented into an R package PPforest. Some extensions of the individual trees in PPF are explored to make the classifier more flexible, to tackle more complex problems, while maintaining interpretability.