论文部分内容阅读
分析了当前一般属性约减算法的特点,给出了广义汉明距离、广义汉明距离矩阵、决策表最小广义汉明距的定义,在此基础上提出了一种保持决策表最小广义汉明距的属性约简算法。该算法对影响决策表最小广义汉明距的关键属性不进行约简,并约去非关键属性,从而有效保持了决策规则的抗噪声能力。仿真实验结果表明:与一般属性约减算法相比,保持决策表最小广义汉明距的属性约简算法的误识率显著下降。
The characteristics of the general reduction algorithm are analyzed. The definitions of the generalized Hamming distance, the generalized Hamming distance matrix and the minimum generalized hamming distance in the decision table are given. Based on this, Attribute reduction algorithm from the distance. The algorithm does not reduce the key attributes that affect the minimum generalized hamming distance of the decision table, and about non-critical attributes, so as to effectively maintain the anti-noise ability of decision rules. The simulation results show that, compared with the general reduction algorithm, the error reduction rate of the attribute reduction algorithm with the smallest generalized hamming distance in the decision table decreases significantly.