论文部分内容阅读
将决策表属性区分能力大小与HU差别矩阵结合起来,建立起属性区别能力与该属性在HU差别矩阵中出现次数之间的关系.提出与HU差别矩阵属性约简相适应,只依赖于等价类基数计算的属性区分能力计算公式,得到HU差别矩阵约简的属性区分能力判定定理.定义了以属性区分能力为基础的属性相对重要性概念,提出以相对重要性为启发式信息的HU属性约简算法.由于该算法不必构造差别矩阵,只依赖于等价类基数的计算,从而大大提高了算法效率.数值算例和实验结果表明,该算法更有利于最优或次优约简结果的搜索.
The relationship between attribute discrimination ability and the number of occurrences of this attribute in the HU discernibility matrix is established by combining the discriminative ability of decision table attributes with the HU discernibility matrix.The proposed method is adaptive to attribute reduction of HU discernibility matrix and only depends on the equivalence The attribute differentiation ability calculation formula of the base cardinality is obtained and the attribute differentiation ability determination theorem of the HU difference matrix reduction is obtained.The concept of the relative importance of attributes based on the attribute differentiation ability is defined and the HU attribute with the relative importance as the heuristic information Reduction algorithm.As the algorithm does not have to construct the discernibility matrix, it only relies on the calculation of equivalence class base, which greatly improves the efficiency of the algorithm.The numerical examples and experimental results show that the algorithm is more conducive to the optimal or suboptimal reduction results Search.