郑州轻工业学院学报（自然科学版）
鄭州輕工業學院學報（自然科學版）
정주경공업학원학보（자연과학판）
JOURNAL OF ZHENGZHOU INSTITUTE OF LIGHT INDUSTRY(NATURAL SCIENCE)
2015年
1期
50-54
,共5页
粗糙集%ID3算法%决策树%相对泛化%等价关系
粗糙集%ID3算法%決策樹%相對汎化%等價關繫
조조집%ID3산법%결책수%상대범화%등개관계
rough set%ID3 algorithm%decision tree%relative generalization%equivalent relationship
针对ID3等传统的决策树算法通常采用单个属性作为分枝判断依据，导致生成树的规模大、形成的规则较难理解的问题，提出了一种以多变量作为结点属性判断条件的算法。该算法利用粗糙集中属性依赖的特性，选择信息系统中条件属性相对决策属性的核属性作为多变量结点属性，使用相对泛化的概念辅助分枝过程，进而生成多变量决策树。通过实例分析与传统的ID3算法进行比较，证明了改进算法的高效性。
針對ID3等傳統的決策樹算法通常採用單箇屬性作為分枝判斷依據，導緻生成樹的規模大、形成的規則較難理解的問題，提齣瞭一種以多變量作為結點屬性判斷條件的算法。該算法利用粗糙集中屬性依賴的特性，選擇信息繫統中條件屬性相對決策屬性的覈屬性作為多變量結點屬性，使用相對汎化的概唸輔助分枝過程，進而生成多變量決策樹。通過實例分析與傳統的ID3算法進行比較，證明瞭改進算法的高效性。
침대ID3등전통적결책수산법통상채용단개속성작위분지판단의거，도치생성수적규모대、형성적규칙교난리해적문제，제출료일충이다변량작위결점속성판단조건적산법。해산법이용조조집중속성의뢰적특성，선택신식계통중조건속성상대결책속성적핵속성작위다변량결점속성，사용상대범화적개념보조분지과정，진이생성다변량결책수。통과실례분석여전통적ID3산법진행비교，증명료개진산법적고효성。
The traditional decision tree algorithms such as ID3 usually uses a single attribute as the basis of branching judgment.The scale of the tree generated by ID3 is very large and rules formed are difficult to understand.Aiming at the problems described above,an algorithm was proposed using multi-variable as the judging conditions of node attributes.By using the property of attribute dependency in rough set and choo-sing nuclear properties of condition attributes relative to decision attributes in the information system as multi-variable node attributes,the algorithm used the concept of relative generalization to aid the branching process and generated a multi-variable decision tree.Through the analysis of example and by comparing with the conventional ID3 algorithm,the high efficiency of the improved algorithm was verified.
