An algorithm of decision tree construction based on attribute support degree

Decision tree algorithms are widely used in data mining and classification systems, because of theirs faster speed, higher accuracy and easier structures. The key to constructing a good decision tree lies in the reasonable choice of attributes. Based on rough set theory and granular computing theory, the paper proposes a concept of attribute support degree to select attributes, using the concept a novel decision tree construction algorithm is presented. The results of experiments on the UCI dataset show that, the decision tree constructed by the new approach tend to have better classification accuracy and stability than ID3 and C4.5.