RST in Decision Tree Pruning

Pruning decision trees is an effective way to overwhelm over-fitting in practice. Various pruning methods have been proposed in many literatures. Though these methods prune decision trees in the light of the principle of ´Minimum Description Length´, they fail to explicitly take into account the impacts of tree scales in the pruning process. This paper proposes a simple decision tree pruning method based on RST (Rough Set Theory). Depth-fitting ratio is introduced for pruning a constructed decision tree, which involves both the depth and the explicit degrees of the sub-trees under evaluation. Experiments on some open data sets shows the feasibility of the new pruning method.