基于决策类划分的新型多变量决策树算法

A New Multivariate the Decision Tree Algorithm Based on the Decision Classification

基于不可分辨关系、复合运算、集合运算和逻辑运算等集合论概念,构造一种新型的多变量决策树算法。该算法包括5个步骤:依据决策属性值划分出决策类;利用决策类之间条件属性集相交判断二义性条件属性值;利用决策类各条件属性值域的不同判断独立决策条件属性值;利用决策类自身条件属性集进行复合运算,获得多变量决策方法;使用或运算符(∨)连接各个部分的决策规则以取得完整的决策规则。以决策树典型训练集(气象信息系统)为例进行验证,其结果表明,该算法行之有效。通过时间复杂度的分析结果表明,该算法较之粗糙集算法更优,而且不亚于ID3算法。

A new multivariate decision tree algorithm is proposed based on the conceptions of indiscernibility relation, compound operations, set operations and logical operation in set theory. It consists of five steps. Firstly, decision is classed according to the deci- sion attribute value. Secondly, the attribute value is worked out according to the set intersection of the decision classes condition attrib- ute. Thirdly, the independent decision condition attribute values are obtained according to each condition attribute field of the decision classification. Fourthly, multivariable decision rules are find out with compound operations of the decision classe condition sets. Final- ly, the OR operation is used to connect every decision rule in order to produce a complete rule. Experiment results demonstrate that the algorithm is effective. The time complexity results demonstrate that this algorithm has better time performance than rough sets algo- rithms, and no less than ID3 algorithm indeed.