论文部分内容阅读
差别矩阵非空元素的个数,直接影响基于差别矩阵的Rough集属性约简算法的效率。分析了几种差别矩阵的不足,基于此,重新定义了一种差别矩阵,该差别矩阵把划分U/C={[x1]C,[x2]C,…,[xn]C}的一个等价类看成一条规则参与区分,从而大大减少了差别矩阵非空元素的个数,提高了Rough集属性约简算法的效率。给出了这几种差别矩阵非空元素的计算公式及其相关定理。提出了一种带启发式知识的约简算法,该算法在很大程度上能找到决策表的最小属性约简。最后给出了对UCI一些数据库的仿真结果。
The number of non-empty elements in the discernibility matrix directly affects the efficiency of the Rough set attribute reduction algorithm based on the discernibility matrix. Based on this, a discernibility matrix is redefined to distinguish one of U / C = {[x1] C, [x2] C, ..., [xn] C} The valence class is regarded as a rule to take part in the distinction, thus greatly reducing the number of non-empty elements in the discernibility matrix and improving the efficiency of Rough set attribute reduction algorithm. The formulas for calculating the non-empty elements of these discernable matrices and their related theorems are given. A reduction algorithm with heuristic knowledge is proposed, which can find the minimum attribute reduction of decision table to a great extent. Finally, some simulation results of UCI database are given.