Can any one tell be about the effect of changing %age Split on decision tree (ID3 and J48) Using Weka?
Pl. reply if any one have idea of this.
The split is about partitioning the dataset you provide into a training and a test set.
The classifier is trained on the train set and subsequently evaluated on the test set.
The larger the training set, the more accurate the classifier usually - trained on more information. But the less informative is the score you get on the test set, since a small test set does not give conclusive results.
This is generic Machine Learning knowledge and applies to decision trees as well as to any other supervised classifier.
Percentage split, partition data for training and testing purposes. Value for percentage may varied on application. When large number of samples available for training, model can be more general to data not present in training set. However, number of training examples is only one factor effect on testing error.
- Badges
- Science topic