A self-learning algorithm for decision tree pre-pruning

Decision tree learning is one of the most widely used machine learning methods. Its two major parts are creating a tree and controlling its size. The advantage of rough set theory for processing uncertain data is used in this paper. From the viewpoint of the certainty of a decision table, the global certainty influenced by each of its condition attributes is used to select split-attributes and control the growing of decision tree. This simplifies the learning process, and solves the problem that the threshold for controlling the growing of decision trees must be given by domain experts in its pre-pruning process. The experimental results show that the method is efficient.