Decision Tree Learning for Drools

Decision trees can be used to represent a large number of expert system rules in a compact way. We describe machine learning algorithms for learning decision trees. We have implemented the algorithms, including bagging and boosting techniques. We have deployed the algorithms in the context of the JBoss Drools rule engine. We present experimental results evaluating the impact of bagging and boosting techniques on the classification accuracy and sizes of trees for several publicly available data sets.