论文部分内容阅读
在粗糙集理论中,最小属性约简未必是对应决策规则数最少的约简.为了从属性数和规则数两个维度消除数据表的冗余信息,提出一种以最少提取规则数和最少属性个数同时为优化目标的多目标属性约简问题及其相应的多目标遗传进化求解算法.该算法在NSGA2(Non-dominated Sorting Genetic Algorithm 2)算法的基础上,以多目标之间的支配关系确定种群个体优先级,并有针对性地引入了精英保留策略、分散进化策略和去重策略.实验结果表明,该算法能够有效地求解本文提出的多目标属性约简问题,其中的分散进化策略、去重策略和精英保留策略增强了进化种群的多样性和收敛性.与NSGA2算法比较,本文算法能获得更多的非支配多目标约简,具有更优的求解能力.
In the rough set theory, the minimum attribute reduction is not necessarily the reduction corresponding to the least number of decision rules.In order to eliminate the redundant information of the data table from the two dimensions of the number of attributes and the number of rules, a method is proposed which uses the least number of rules and the least attributes The number is also the multi-objective attribute reduction problem of the optimization objective and its corresponding multi-objective genetic evolution algorithm.On the basis of Non-dominated Sorting Genetic Algorithm 2 (NSGA2) algorithm, this algorithm takes the multi-objective dominance relationship The priority of individual population is determined, and elite retention strategy, decentralized evolution strategy and deduplication strategy are introduced.Experimental results show that this algorithm can effectively solve the multi-objective attribute reduction problem proposed in this paper, in which the decentralized evolution strategy , De-duplication strategy and elite retention strategy enhance the diversity and convergence of evolutionary population.Compared with NSGA2 algorithm, the proposed algorithm can get more non-dominated multi-objective reduction and has better solving ability.