Publication at Faculty of Mathematics and Physics |
2006
Abstract
The current paper uses the model of trees with soft splits suggested by Quinlan and implemented in C4.5, but with a substantially different the training algorithm. The method uses simulated annealing, so it is quite computationally expensive.
However, this allows to adjust the soft thresholds in groups of the nodes simultaneously in a way that better captures interactions between several predictors than the original approach. Our numerical test with data derived from an experiment in particle physics shows that besides the expected better approximation of the training data, also smaller generalization error is achieved.