摘要
针对现有判定树算法在处理空缺值和连续值以及知识表达上不精确性和复杂性问题,提出基于云变换和Rough扩展模型的判定树构造算法。该算法利用云变换来离散化连续属性,然后根据概念集,采用极大判定法对每个数值型属性的原始属性值进行软划分,从而得到离散属性值。最后利用特性关系下的加权平均粗糙度来选取当前结点的分裂属性来递归生成判定树。与C5.0算法相比,新算法可妥善处理空缺值、合理离散连续属性。试验结果表明,该算法具有良好的实用性。
Current decision tree cannot handle missing data and continuous data effectively and there exists the complexity and uncertainty in knowledge expression.A new decision tree construction based on cloudtransformation and tough set theory is presented.Firstly,it utilizes cloud transform to discretize continuous data.Then,Based on the obtained concept and adopting method of maximum determinant,the thesis gets the ultimate discretize data by soft-classifying the original attribute data of every numerical attribute.Finally,it generates decision tree recursively by selecting the splitting attribute of the current nodes according to the weighted mean roughness under the characteristic relation-based rough sets.The new algorithm can handle incomplete data and discretize continuous data more effectively than C5.0.〗The experiment results show that this method is feasible.
出处
《计算技术与自动化》
2010年第3期140-144,共5页
Computing Technology and Automation
基金
湖南教育厅科学研究基金项目(08C248)
湖南教育厅科学研究基金项目(09C297)
关键词
云变换
粗糙集
离散属性
超熵
cloud transformation
rough set
category attributes
hyper entropy
