基于可行域的遗传约简算法

Genetic Reduction Algorithm Based on Feasible Region

在已有的遗传属性约简算法的基础上,通过引入约简的可行域概念,提出了基于可行域的遗传约简算法.可行域保持系统的分类能力,缩小了原问题的搜索空间,进而减小了问题的复杂度,适应度函数中引入与互信息相关的惩罚因子保证了算法在可行域中搜索.实验结果表明该算法既克服了启发性算法的缺陷,较之已有的基于遗传算法的约简算法也有效率改进.

Based on the known genetic attribute reduction algorithms, by introducing the concept of feasible region, this paper proposes a genetic reduction algorithm based on feasible region. Feasible region maintains the ability of classification of the system, and it narrows the search space of original question, then reduces the burden of calculation. The punish factor related with mutual information is introduced in Fitness function, so it can keep algorithm search in the feasible region, experiments results showed the algorithm can overcome the shortcoming of heuristic attribute reduction, compared with the known genetic attribute reduction algorithms it also have the advantage of efficiency.