基于变精度粗糙集模型的属性约简方法研究

提出了一种对存在噪声和不完整数据的决策系统在变精度粗糙集模型下进行属性最小相对约简的方法,将由属性对分类的影响程度和β近似精度共同定义的属性重要性度量作为启发式信息引入遗传算法,通过修正操作算子修复个体,使得个体所对应的属性子集的分类能力不变;修正操作算子中对各属性的属性重要性使用贪心策略进行局部寻优.对遗传算法的各操作算子进行优化,既保证遗传操作过程中种群的多样性,又保证算法能快速收敛.最后通过实例验证了算法的有效性.

Study on Algorithm for Reduction of Attributes Based on Varied Precision Rough Set

A genetic algorithm, which is based on varied precision rough set, for minimizing relative reduction on a decision system which includes noises and non - integrated data is proposed. The significance of attributes defined from the influence of an attribute on classifying and 13 approximate precision is regarded as heuristic information, which is introduced into genetic algorithm. A modify operator, which uses greedy strategy to select the most significant attribute, is used to repair the individuals in the colony in order to maintain the ability of classification. The operators of genetic algorithm are optimized, and this ensures both the diversity among the individuals in the colony and the rapid convergence of the algorithm. The validity of the algorithm in this paper is proved by an instance.