Are there any free Java libraries available for training decision trees that allow for the setting of a separate validation set for reduced error pruning?
For Weka (what I'm currently using), it only allows for n-fold cross validation using a random subset of the data. However, as I am working with time series data it would be beneficial to use the first half of my data as the training data, and use the second half (future data) for validation/pruning.
With n-fold validation, overfitting is a serious problem and is leading to barely above ~50% accuracy.
