一种新的基于区分矩阵的值约简算法

摘要

在对基于区分矩阵的值约简算法进行了分析发现，这些算法在区分矩阵构造时或数据处理时，忽略了对于同在一个决策属性类对象集包含的信息， BDMVR正是利用了这些被忽略的信息，构造出新的启发信息用于指导值约简算法。新算法更充分地利用了区分矩阵中包含的信息，且与对象和条件属性顺序无关，都能得到相同规则。通过算例分析和在5个数据集上的比对实验证明了算法的有效性。%Analyzed on the value reduction algorithm based on discernibility matrix, these algorithms ignored the information contain in object set of the same decision attribute class when the discernibility matrix structure or data processing. BDMVR uses these neglected information, construct a new heuristic information used to guide value reduction algorithm. The new algorithm make full use of the information contained in the discernibility matrix, it has nothing to do with the objects and condition attributes order and can get the same rules. Through the example analysis and the comparison experiments on five data-sets prove the effectiveness of the algorithm.