关键词: 粗糙集, 决策树, 粒分辩关系

Abstract: Aiming at the problem of inaccurate constructing a decision tree based on the information gain algorithm to determine the preferred attributes and an algorithm of the rough set using core attributes as the preferred attributes, in granulation theory, according to its nature value the attribute is divided into several basic granulation. Through conditional length（the volume of granulation resolution） and the resolution of granulation relativing to decision granulation（the class of resolution）, the attribute is divided. Multi-attribute selection problem is solved by using the simple formula. Theory and example analysis results show that the algorithm achieves precision and formula is simple effective and the temporal and spatial complexity is low.

Key words: rough set, decision tree, granulatio relation