RoughTree A Classifier with Naive-Bayes and Rough Sets Hybrid in Decision Tree Representation
Yangsheng Ji, Lin Shang · 2007 IEEE International Conference on Granular Computing (GRC 2007) · 2007
This paper presents a semi-naive classifier named RoughTree, which is designed to alleviate the attribute interdependence problem of Naive Bayesian classifier. RoughTree uses the attribute dependence detecting measure in rough sets and splits the dataset into subspaces according to the selected attributes, which hold the maximum values by the attribute dependence measure. This process continues the same way a decision tree splits until the stopping criterion is satisfied. Then, the result is a tree-like model and each leaf in the RoughTree is replaced by a Naive-Bayesian classifier. RoughTree eliminates the attribute dependences in its leaves and the experimental results show that RoughTree can achieve better performance than Naive Bayesian classifier.