基于粗集的多变量决策树方法研究

A Multivariate Decision Tree Method Based on Rough Set

为了使构造的决策树更简单,规则更容易被理解且精度更高,文章基于粗糙集理论提出了一种对属性约简及泛化的多变量决策树算法。该方法采用条件属性的加权平均粗糙度这个指标来选择测试属性构造决策树。实验表明该方法较ID3算法得到的决策树更小且分类准确率更高。文章还展望用核属性以外的条件组合属性作测试属性构造更简化的多变量决策树。

The paper presents a multivariable Decision Tree Algorithm on the use of attribute reduction and generalization based on rough set theory, for a easier and higher accuracy tree, especially using the indicator that weighted average roughness to select testing attributes and to build decision tree. Experiments show that this method makes more thee and higher classification accuracy rate than ID3. Furthermore, this paper also prospects use multi-conditional attributes other than nuclear attributes as testing attributes to build more simplified multivariable decision tree.