摘要
针对传统的基于区分矩阵属性约简算法在构建区分矩阵时生成很多冗余元素,空间和时间性能上不够理想的缺点,利用动态数组及新的约简过程进一步改进,最后把新的属性约简算法应用到决策树的构建上,实验证明,该算法能够提高决策树分类性能。
Aimed at these shortcomings, the traditional attribute reduction algorithm based on discemability matrix generated a lot of redundant elements when build discernability matrix, both space and time performance were not satisfactory, It was given further improvement by using dynamic array and a new reduction process, finally, applied the new attribute reduction algorithm to the decision tree building. The experimental study showed that this algorithm could greatly enhance the performance of decision tree classification.
出处
《铁路计算机应用》
2009年第4期11-14,共4页
Railway Computer Application
关键词
粗糙集
属性约简
决策树
区分矩阵
rough set
attribute reduction
decision tree
discernability matrix
