A Heuristic Partition Method of Numerical Attributes in Classification Tree Construction

• Main Authors: Wen-Ke Tseng, 曾文科
• Other Authors: Ester Yen
• Format: Others
• Language: en_US
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/64229514474119881705

碩士 === 大同大學 === 資訊經營學系(所) === 92 === Inductive Learning, a kind of learning methods, has been applied extensively in Machine Learning. Thus, Classification tree is a well-known method in Inductive Learning. The ID3, a popular classification tree algorithm, had been proposed by Quinlan on 1986. Quinlan proposed the C4.5 algorithm on 1993 again. The C4.5 has not been efficiently searching the splitting points on numerical attributes. Therefore, some researchers had proposed improved approaches and new partition methods for the partition on numerical attributes. However, these approaches and methods have its assumptions and restrictions. So we have proposed a heuristic partition method to improve the defect, which the C4.5 algorithm could not process numerical attributes efficiently. Since the heuristic partition method is based on C4.5 algorithm, the method can greatly reduce the time for searching splitting point on numerical attributes.